<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.1 20151215//EN" "https://jats.nlm.nih.gov/publishing/1.1/JATS-journalpublishing1.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="review-article" dtd-version="1.1" specific-use="sps-1.9" xml:lang="pt">
  <front>
    <journal-meta>
      <journal-id journal-id-type="nlm-ta">Arq Bras Cardiol</journal-id>
      <journal-id journal-id-type="publisher-id">abc</journal-id>
      <journal-title-group>
        <journal-title>Arquivos Brasileiros de Cardiologia</journal-title>
        <abbrev-journal-title abbrev-type="publisher">Arq. Bras. Cardiol.</abbrev-journal-title>
      </journal-title-group>
      <issn pub-type="ppub">0066-782X</issn>
      <issn pub-type="epub">1678-4170</issn>
      <publisher>
        <publisher-name>Sociedade Brasileira de Cardiologia - SBC</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="publisher-id" specific-use="scielo-v3">VNQtkGdM85HdjCXKJ6ZgXkQ</article-id>
      <article-id pub-id-type="publisher-id" specific-use="scielo-v2">S0066-782X2020000400718</article-id>
      <article-id pub-id-type="publisher-id">abc.20180431</article-id>
      <article-id pub-id-type="doi">10.36660/abc.20180431</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Artigo de Revis&#227;o</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Intelig&#234;ncia Artificial em Cardiologia: Conceitos, Ferramentas e Desafios &#8211; &#8220;Quem Corre &#233; o Cavalo, Voc&#234; Precisa ser o J&#243;quei&#8221;</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0381-3344</contrib-id>
          <name>
            <surname>Souza</surname>
            <given-names>Erito Marques de</given-names>
            <suffix>Filho</suffix>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Fernandes</surname>
            <given-names>Fernando de Amorim</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Soares</surname>
            <given-names>Celine Lacerda de Abreu</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Seixas</surname>
            <given-names>Flavio Luiz</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Santos</surname>
            <given-names>Alair Augusto Sarmet M.D. dos</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Gismondi</surname>
            <given-names>Ronaldo Altenburg</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Mesquita</surname>
            <given-names>Evandro Tinoco</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-1466-9413</contrib-id>
          <name>
            <surname>Mesquita</surname>
            <given-names>Claudio Tinoco</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
      </contrib-group>
      <aff id="aff1">
        <label>1</label>
        <institution content-type="orgname">Universidade Federal Fluminense</institution>
        <addr-line>
          <named-content content-type="city">Niter&#243;i</named-content>
          <named-content content-type="state">RJ</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Universidade Federal Fluminense, Niter&#243;i, RJ &#8211; Brasil</institution>
      </aff>
      <aff id="aff2">
        <label>2</label>
        <institution content-type="orgname">Universidade Federal Rural do Rio de Janeiro</institution>
        <institution content-type="orgdiv1">Departamento de Tecnologias e Linguagens</institution>
        <addr-line>
          <named-content content-type="city">Nova Igua&#231;u</named-content>
          <named-content content-type="state">RJ</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Universidade Federal Rural do Rio de Janeiro - Departamento de Tecnologias e Linguagens, Nova Igua&#231;u, RJ &#8211; Brasil</institution>
      </aff>
      <author-notes>
        <corresp id="c01"> Correspond&#234;ncia: Erito Marques de Souza Filho &#8226; Universidade Federal Fluminense - Departamento de Medicina Cl&#237;nica &#8211; Av. Marques do Paran&#225;, 303. CEP 24033-900, Niter&#243;i, RJ &#8211; Brasil E-mail: <email>mederitomarques@gmail.com</email>
				</corresp>
        <fn fn-type="con">
          <p>Contribui&#231;&#227;o dos autores</p>
          <p>Concep&#231;&#227;o e desenho da pesquisa: Souza Filho EM, Seixas FL, Santos AASMD, Gismondi RA, Mesquita ET, Mesquita CT; Obten&#231;&#227;o de dados: Souza Filho EM, Fernandes FA, Soares CLA, Santos AASMD; An&#225;lise e interpreta&#231;&#227;o dos dados: Souza Filho EM, Fernandes FA, Soares CLA, Seixas FL, Gismondi RA, Mesquita ET, Mesquita CT; Reda&#231;&#227;o do manuscrito: Souza Filho EM, Fernandes FA, Soares CLA, Gismondi RA, Mesquita ET; Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante: Seixas FL, Santos AASMD, Mesquita CT.</p>
        </fn>
        <fn fn-type="conflict">
          <p>Potencial conflito de interesses</p>
          <p>Declaro n&#227;o haver conflito de interesses pertinentes.</p>
        </fn>
      </author-notes>
      <pub-date date-type="pub" publication-format="electronic">
        <day>29</day>
        <month>05</month>
        <year>2020</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>04</month>
        <year>2020</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>04</month>
        <year>2020</year>
      </pub-date>
      <volume>114</volume>
      <issue>4</issue>
      <fpage>718</fpage>
      <lpage>725</lpage>
      <history>
        <date date-type="received">
          <day>21</day>
          <month>12</month>
          <year>2018</year>
        </date>
        <date date-type="rev-recd">
          <day>16</day>
          <month>07</month>
          <year>2019</year>
        </date>
        <date date-type="accepted">
          <day>28</day>
          <month>08</month>
          <year>2019</year>
        </date>
      </history>
      <permissions>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/" xml:lang="en">
          <license-p> This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License, which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. </license-p>
        </license>
      </permissions>
      <abstract>
        <title>Resumo</title>
        <p>Os recentes avan&#231;os ao n&#237;vel de hardware e a crescente exig&#234;ncia de personaliza&#231;&#227;o dos cuidados associados &#224;s necessidades urgentes de cria&#231;&#227;o de valor para os pacientes contribu&#237;ram para que a Intelig&#234;ncia Artificial (IA) promovesse uma mudan&#231;a significativa de paradigma nas mais diversas &#225;reas do conhecimento m&#233;dico, em particular em Cardiologia, por sua capacidade de apoiar a tomada de decis&#245;es e melhorar o desempenho diagn&#243;stico e progn&#243;stico. Nesse contexto, o presente trabalho faz uma revis&#227;o n&#227;o-sistem&#225;tica dos principais trabalhos publicados sobre IA em Cardiologia, com foco em suas principais aplica&#231;&#245;es, poss&#237;veis impactos e desafios.</p>
      </abstract>
      <kwd-group xml:lang="pt">
        <kwd>Intelig&#234;ncia Artificial/tend&#234;ncias</kwd>
        <kwd>Sistemas de Computa&#231;&#227;o/tend&#234;ncias</kwd>
        <kwd>Aprendizado de M&#225;quina/tend&#234;ncias</kwd>
        <kwd>Doen&#231;as Cardiovasculares</kwd>
        <kwd>Tomada de Decis&#227;o Cl&#237;nica</kwd>
      </kwd-group>
      <counts>
        <fig-count count="4"/>
        <table-count count="2"/>
        <equation-count count="0"/>
        <ref-count count="59"/>
        <page-count count="8"/>
      </counts>
    </article-meta>
  </front>
  <body>
    <sec sec-type="intro">
      <title>Introdu&#231;&#227;o</title>
      <p>A vida cotidiana de uma pessoa exige uma enorme quantidade de conhecimento sobre o mundo e o volume de dados em sa&#250;de cresce exponencialmente em todo o mundo.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> Por outro lado, o conhecimento biom&#233;dico est&#225; sempre se expandindo de maneira ativa e din&#226;mica e n&#227;o pode ser processado ou armazenado por um &#250;nico c&#233;rebro humano. Essa situa&#231;&#227;o torna muito dif&#237;cil para o m&#233;dico contempor&#226;neo manter-se atualizado com um amplo espectro de novos dados e descobertas, bem como em rela&#231;&#227;o &#224; utiliza&#231;&#227;o dessas informa&#231;&#245;es com facilidade e em tempo h&#225;bil.<sup><xref ref-type="bibr" rid="B2">2</xref></sup> Acrescentam-se a esse quadro as significativas taxas de <italic>burnout</italic> entre profissionais de sa&#250;de<sup><xref ref-type="bibr" rid="B3">3</xref> , <xref ref-type="bibr" rid="B4">4</xref></sup> e o importante impacto de erros m&#233;dicos &#8211; que nos Estados Unidos representam a terceira principal causa de morte.<sup><xref ref-type="bibr" rid="B5">5</xref></sup> Esse panorama traz consigo a necessidade de reorganizar a estrutura produtiva dos servi&#231;os de sa&#250;de, associada a v&#225;rios desafios e novas perspectivas. Dado que o sistema de sa&#250;de atual &#233; geralmente improdutivo e /ou caro, &#233; imperativo desenvolver estrat&#233;gias alternativas e inovadoras. O foco central para alcan&#231;ar esse objetivo deve ser aumentar o valor para o paciente &#8211; resultados alcan&#231;ados por d&#243;lar gasto &#8211; para que bons resultados, obtidos com efici&#234;ncia, sejam um objetivo a ser perseguido.<sup><xref ref-type="bibr" rid="B6">6</xref></sup></p>
      <p>Al&#233;m disso, os recentes avan&#231;os ao n&#237;vel de hardware relacionados ao processamento paralelo, a exist&#234;ncia de v&#225;rios m&#233;todos de aprendizado de m&#225;quina e a enorme quantidade de dados anotados contribu&#237;ram para que a intelig&#234;ncia artificial (IA) promovesse uma mudan&#231;a significativa de paradigma nas mais diversas &#225;reas do conhecimento m&#233;dico, e particularmente em Cardiologia, por sua capacidade de apoiar a tomada de decis&#245;es que podem melhorar o desempenho diagn&#243;stico e progn&#243;stico. Esses impactos devem ser avaliados na perspectiva da seguran&#231;a do paciente, personaliza&#231;&#227;o dos cuidados, cria&#231;&#227;o de valor para os pacientes, no &#226;mbito da vigil&#226;ncia tecnol&#243;gica &#8211; que gradualmente consolida a IA como fundamental para uma pr&#225;tica m&#233;dica de excel&#234;ncia.<sup><xref ref-type="bibr" rid="B7">7</xref> - <xref ref-type="bibr" rid="B11">11</xref></sup></p>
      <p>Esse cen&#225;rio faz com que a IA, dada sua import&#226;ncia, seja considerada por muitos como a nova eletricidade. As principais Revistas de Cardiologia publicaram revis&#245;es nessa &#225;rea e o n&#250;mero de artigos sobre o assunto segue uma tend&#234;ncia crescente, como mostra a <xref ref-type="fig" rid="f01">Figura 1</xref> &#8211; esse comportamento tamb&#233;m &#233; observado em outras especialidades m&#233;dicas, como a neurologia. Portanto, o presente trabalho faz uma revis&#227;o n&#227;o-sistem&#225;tica dos principais trabalhos publicados sobre IA em Cardiologia, com foco em suas principais aplica&#231;&#245;es, potenciais impactos e desafios. A pr&#243;xima se&#231;&#227;o apresenta os fundamentos conceituais do t&#243;pico, seguidos do motivo pelo qual a cardiologia precisa da IA e de suas principais ferramentas. Por fim, s&#227;o apresentados os principais desafios, perspectivas e conclus&#245;es.</p>
      <p>
        <fig id="f01">
          <label>Figura 1</label>
          <caption>
            <title>&#8211; <italic>Evolu&#231;&#227;o do n&#250;mero de trabalhos relacionados (Intelig&#234;ncia Artificial ou Aprendizado de M&#225;quina, ou Aprendizado Autom&#225;tico) e Cardiologia. Fonte: Pubmed. Acessado em 15/12/2018. Palavras do Medical Subject Headings (MeSH): Cardiologia e Aprendizado de M&#225;quina.</italic></title>
          </caption>
          <graphic xlink:href="1678-4170-abc-114-04-0718-80431-gf01.jpg"/></fig>
      </p>
      <sec>
        <title>O que &#233; intelig&#234;ncia artificial (IA)?</title>
        <p>O termo AI foi utilizado pela primeira vez na Confer&#234;ncia de Dartmouth em 1956.<sup><xref ref-type="bibr" rid="B12">12</xref></sup> No entanto, a possibilidade de que as m&#225;quinas fossem capazes de simular o comportamento humano e realmente pensar foi levantada anteriormente por Alan Turing em 1950, que desenvolveu um teste para diferenciar os seres humanos de m&#225;quinas &#8211; denominado teste de Turing.<sup><xref ref-type="bibr" rid="B13">13</xref></sup></p>
        <p>Basicamente, a IA &#233; o produto da combina&#231;&#227;o de modelos matem&#225;ticos sofisticados e computa&#231;&#227;o, que permite o desenvolvimento de algoritmos complexos capazes de emular a intelig&#234;ncia humana. Todo esse processo inicia-se com a constru&#231;&#227;o de um banco de dados representativo do problema que se deseja estudar &#8211; coletado e processado adequadamente &#8211; chamados dados saud&#225;veis. Essa etapa &#233; de fundamental import&#226;ncia, pois os algoritmos provavelmente n&#227;o ter&#227;o um bom desempenho se esse pr&#233;-requisito n&#227;o for obtido: &#8220; <italic>lixo na entrada, lixo na sa&#237;da&#8221;</italic> .</p>
        <p>A natureza desses dados &#233; bastante variada, variando de dados socioambientais, cl&#237;nico-laboratoriais a dados &#244;micos (por exemplo, metaboloma, proteoma, epigenoma, lipidoma) e informa&#231;&#245;es das intensidades de vermelho, verde e azul (sistema RGB) de cada pixel que comp&#245;e uma imagem, por exemplo. Fontes igualmente diversificadas de tais dados incluem aquelas obtidas de registros m&#233;dicos eletr&#244;nicos ou mesmo dispositivos &#8220; <italic>wearable</italic> &#8221; (&#8220;vest&#237;veis&#8221;). Nesse contexto, o termo &#8220;Big Data&#8221; &#233; utilizado para descrever uma enorme cole&#231;&#227;o de dados para os quais os m&#233;todos tradicionais de an&#225;lise n&#227;o s&#227;o bem-sucedidos na an&#225;lise, pesquisa, interpreta&#231;&#227;o e armazenamento.<sup><xref ref-type="bibr" rid="B9">9</xref></sup></p>
        <p>Destacamos o uso dessas ferramentas em problemas de classifica&#231;&#227;o, regress&#227;o e clusteriza&#231;&#227;o. Depois de obter dados saud&#225;veis e construir o banco de dados, &#233; importante avaliar quais modelos matem&#225;ticos de IA s&#227;o mais apropriados para o problema que se deseja resolver. Em seguida, os modelos escolhidos devem ser implementados utilizando alguma linguagem de programa&#231;&#227;o.</p>
        <p>Uma combina&#231;&#227;o de modelos tamb&#233;m pode ser &#250;til. Os resultados obtidos pelo algoritmo devem ser analisados em termos de coer&#234;ncia e adequa&#231;&#227;o. Essas etapas est&#227;o resumidas na <xref ref-type="fig" rid="f02">Figura 2</xref> .</p>
        <p>
          <fig id="f02">
            <label>Figura 2</label>
            <caption>
              <title>&#8211; <italic>Ilustra&#231;&#227;o principal.</italic></title>
            </caption>
            <graphic xlink:href="1678-4170-abc-114-04-0718-80431-gf02.jpg"/></fig>
        </p>
      </sec>
      <sec>
        <title>Por que a cardiologia precisa de intelig&#234;ncia artificial?</title>
        <p>O desenvolvimento de algoritmos de IA tem a vantagem de n&#227;o exigir muitas suposi&#231;&#245;es em rela&#231;&#227;o aos dados subjacentes.<sup><xref ref-type="bibr" rid="B8">8</xref></sup> Outro ponto &#233; que a natureza desses modelos matem&#225;tico-computacionais permite, a partir de dados observacionais, um alto n&#237;vel de evid&#234;ncia devido ao seu alto desempenho, o que certamente representa uma mudan&#231;a significativa de paradigma na medicina baseada em evid&#234;ncias. Deve-se notar que os ensaios cl&#237;nicos tradicionais s&#227;o geralmente lentos, caros, demorados e limitados em tamanho.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> Al&#233;m disso, quando o banco de dados &#233; alimentado com mais dados (saud&#225;veis), em geral, h&#225; uma melhoria no desempenho dos algoritmos &#8211; o que permite que os estudos tenham um car&#225;ter cont&#237;nuo ao longo do tempo.</p>
        <p>Esse novo arqu&#233;tipo pode guiar a aloca&#231;&#227;o de recursos escassos na &#225;rea da sa&#250;de e facilitar a identifica&#231;&#227;o eficiente e precisa de decis&#245;es que favore&#231;am a individualiza&#231;&#227;o do cuidado com base no fluxo de informa&#231;&#245;es que emergem de um ecossistema integrado e complexo: &#233; uma medicina de precis&#227;o.<sup><xref ref-type="bibr" rid="B15">15</xref> , <xref ref-type="bibr" rid="B16">16</xref></sup> Portanto, pode-se inferir que a pr&#225;tica das ci&#234;ncias cardiovasculares ter&#225; impactos significativos, que se traduzir&#227;o em uma abordagem personalizada e melhores resultados.</p>
      </sec>
      <sec>
        <title>Conceitos b&#225;sicos em intelig&#234;ncia artificial</title>
        <p>Um banco de dados gen&#233;rico pode ser organizado em uma matriz de linhas e colunas. Cada linha indica um elemento de um conjunto de objetos a ser avaliado de acordo com as mesmas caracter&#237;sticas. Cada coluna, por sua vez, expressa os valores de um determinado atributo para as v&#225;rias linhas no banco de dados e cada linha representa uma li&#231;&#227;o a ser aprendida pelo modelo matem&#225;tico-computacional. Dessa maneira, o termo Aprendizado de M&#225;quina (ML, <italic>machine learning</italic> ) traz consigo a possibilidade de &#8220;aprender&#8221; a partir de um conjunto de li&#231;&#245;es. O termo IA &#233; frequentemente utilizado de forma intercambi&#225;vel com o termo ML. No entanto, ML &#233; um subconjunto de algoritmos de IA relacionados &#224; capacidade de aprender a partir de uma grande quantidade de dados. A IA &#233; mais ampla e engloba a execu&#231;&#227;o de tarefas que normalmente est&#227;o relacionadas &#224; intelig&#234;ncia humana, tais como reconhecimento de padr&#245;es, resolu&#231;&#227;o de problemas, compreens&#227;o de linguagem ou reconhecimento de objetos e sons.<sup><xref ref-type="bibr" rid="B17">17</xref></sup></p>
        <p>Costuma-se dizer que os tipos de aprendizado podem ser:</p>
        <list list-type="alpha-lower">
          <list-item>
            <p><bold>Supervisionados:</bold> quando o algoritmo recebe informa&#231;&#245;es sobre cada li&#231;&#227;o, bem como os r&#243;tulos associados a ela, tendo um papel importante em rela&#231;&#227;o &#224; previs&#227;o. Por exemplo, se &#233; desejado prever se um paciente &#233; mais suscet&#237;vel &#224; tosse com o uso de inibidores da enzima de convers&#227;o da angiotensina, a an&#225;lise deve ser realizada baseada em um banco de dados saud&#225;veis contendo um grupo de pacientes que demonstrou essa rea&#231;&#227;o e em outro grupo no qual esse fato n&#227;o foi observado.</p>
          </list-item>
          <list-item>
            <p><bold>N&#227;o-supervisionado:</bold> quando os r&#243;tulos das li&#231;&#245;es n&#227;o s&#227;o fornecidos <italic>a priori</italic> , cabe ao algoritmo encontrar estruturas ocultas no banco de dados. Um exemplo hipot&#233;tico &#233; a clusteriza&#231;&#227;o de um banco de dados de pacientes com cardiomiopatia hipertr&#243;fica de acordo com os achados de exame de imagem.</p>
          </list-item>
          <list-item>
            <p><bold>Refor&#231;o:</bold> inspirado na biologia comportamental, &#233; um tipo de aprendizado baseado em recompensa.<sup><xref ref-type="bibr" rid="B18">18</xref> , <xref ref-type="bibr" rid="B19">19</xref></sup></p>
          </list-item>
        </list>
        <p>Outro conceito importante &#233; o da computa&#231;&#227;o cognitiva. Ele pode ser entendido como um conjunto de sistemas de autoaprendizagem destinados a imitar o processo de pensamento humano com base no uso de ferramentas de ML, reconhecimento de padr&#245;es e processamento natural de linguagem.<sup><xref ref-type="bibr" rid="B9">9</xref></sup> O IBM Watson &#233; um exemplo da computa&#231;&#227;o cognitiva na &#225;rea m&#233;dica.<sup><xref ref-type="bibr" rid="B20">20</xref> , <xref ref-type="bibr" rid="B21">21</xref></sup></p>
      </sec>
      <sec>
        <title>Algumas ferramentas e aplicativos de intelig&#234;ncia artificial</title>
        <p>Atualmente, existem v&#225;rios modelos de ML, cada um deles com diversas particularidades, usos e limita&#231;&#245;es variados. As aplica&#231;&#245;es de alguns desses modelos em Cardiologia s&#227;o explicadas nos par&#225;grafos seguintes, enquanto uma breve descri&#231;&#227;o de cada um deles e seu tipo &#233; mostrada na <xref ref-type="table" rid="t1">Tabela 1</xref> .</p>
        <p>
          <table-wrap id="t1">
            <label>Tabela 1</label>
            <caption>
              <title>&#8211; Breve descri&#231;&#227;o e classifica&#231;&#227;o das principais ferramentas de ML</title>
            </caption>
            <table frame="hsides" rules="groups">
              <colgroup width="33%">
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead>
                <tr>
                  <th align="left">Ferramenta</th>
                  <th>Descri&#231;&#227;o</th>
                  <th>Aprendizado</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td>SVM</td>
                  <td align="center">&#201; &#250;til para problemas de classifica&#231;&#227;o de dois grupos. A ideia &#233; encontrar uma fun&#231;&#227;o chamada hiperplano a partir da resolu&#231;&#227;o de um sistema linear constru&#237;do a partir das v&#225;rias li&#231;&#245;es do subconjunto de treinamento. <xref ref-type="bibr" rid="B40"><sup>40</sup></xref> Esse hiperplano &#233; utilizado para agrupar as li&#231;&#245;es do subconjunto de teste em dois grupos disjuntos.</td>
                  <td align="center">Supervisionado</td>
                </tr>
                <tr>
                  <td>NB</td>
                  <td align="center">Foi inspirado nos estudos do reverendo Bayes sobre a probabilidade condicional. <xref ref-type="bibr" rid="B41"><sup>41</sup></xref> Essas probabilidades s&#227;o usadas para identificar a categoria (de um total de n poss&#237;vel) a que uma li&#231;&#227;o espec&#237;fica pertence. <xref ref-type="bibr" rid="B42"><sup>42</sup></xref>
									</td>
                  <td align="center">Supervisionado</td>
                </tr>
                <tr>
                  <td>KNN</td>
                  <td align="center">Diz-se que uma norma de vetor &#233; uma fun&#231;&#227;o matem&#225;tica, que satisfaz propriedades espec&#237;ficas, e associa um vetor a um valor maior ou igual a zero. <xref ref-type="bibr" rid="B43"><sup>43</sup></xref> A norma da diferen&#231;a entre dois vetores &#233; a dist&#226;ncia entre eles. O KNN utiliza uma norma para calcular a dist&#226;ncia entre todos os vetores (li&#231;&#245;es) que comp&#245;em o banco de dados. Ent&#227;o, para cada vetor do banco de dados, os vetores k mais pr&#243;ximos a ele s&#227;o determinados. A inclus&#227;o em um determinado grupo &#233; obtida a partir de um sistema de vota&#231;&#227;o majorit&#225;ria entre os vizinhos. <xref ref-type="bibr" rid="B44"><sup>44</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B45"><sup>45</sup></xref>
									</td>
                  <td align="center">Supervisionado</td>
                </tr>
                <tr>
                  <td>AG</td>
                  <td align="center">Algoritmos inspirados na evolu&#231;&#227;o biol&#243;gica das esp&#233;cies, nos quais cada poss&#237;vel candidato para solucionar o problema &#233; modelado como um cromossomo constitu&#237;do por um conjunto de genes, que durante a execu&#231;&#227;o do algoritmo &#233; submetido a opera&#231;&#245;es de cruzamento e muta&#231;&#227;o para obter melhor solu&#231;&#245;es que as atuais. <xref ref-type="bibr" rid="B46"><sup>46</sup></xref> Dessa forma, eles permitem que um banco de dados seja separado, por exemplo, em dois grupos distintos &#8211; que possuem ou n&#227;o uma caracter&#237;stica espec&#237;fica.</td>
                  <td align="center">Supervisionado</td>
                </tr>
                <tr>
                  <td>RF</td>
                  <td align="center">Esse m&#233;todo &#233; baseado na constru&#231;&#227;o de v&#225;rias &#225;rvores de decis&#227;o. O primeiro passo &#233; obter v&#225;rias amostras aleat&#243;rias (com reposi&#231;&#227;o) de li&#231;&#245;es para criar outros bancos de dados, um processo chamado &#8220;bootstrapping&#8221;. Cada um desses novos bancos de dados dar&#225; origem a uma &#225;rvore de decis&#227;o, obtida de forma iterativa, a partir de um subconjunto de vari&#225;veis (caracter&#237;sticas). Ap&#243;s a constru&#231;&#227;o de todas as &#225;rvores, uma nova li&#231;&#227;o no banco de dados deve ser alocada ao grupo que tem o maior n&#250;mero de &#225;rvores de decis&#227;o, mostrando que ele pertence a esse grupo (maioria dos votos). <xref ref-type="bibr" rid="B47"><sup>47</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B48"><sup>48</sup></xref>
									</td>
                  <td align="center">Supervisionado</td>
                </tr>
                <tr>
                  <td>K-means</td>
                  <td align="center">Permite particionar um banco de dados em k grupos com caracter&#237;sticas semelhantes. Para isso, &#233; necess&#225;rio atualizar, de maneira iterativa, um conjunto de vetores, chamados centroides de refer&#234;ncia de cada grupo e calcular a dist&#226;ncia de cada li&#231;&#227;o para cada um. Uma li&#231;&#227;o &#233; sempre alocada ao centroide para o qual ela tem a menor dist&#226;ncia. O gr&#225;fico de cotovelo &#233; geralmente utilizado para determinar o n&#250;mero ideal de grupos a serem separados do banco de dados. <xref ref-type="bibr" rid="B49"><sup>49</sup></xref>
									</td>
                  <td align="center">N&#227;o supervisionado</td>
                </tr>
                <tr>
                  <td>ANN</td>
                  <td align="center">Inspirado nos sistemas nervosos biol&#243;gicos, &#233; utilizada uma estrutura chamada gr&#225;fico &#8211; um conjunto de n&#243;s e bordas &#8211; em que os n&#243;s s&#227;o estratificados e conectados por bordas com valores, que representam um peso atribu&#237;do a uma determinada conex&#227;o. A ideia &#233; que, a partir de um conjunto de entradas, esses pesos sejam utilizados corretamente para produzir uma sa&#237;da. V&#225;rias arquiteturas foram propostas para redes neurais, desde as mais simples, como a perceptron, at&#233; as mais sofisticadas, como a fun&#231;&#227;o de base radial, redes convolucionais e aprendizado profundo. No aprendizado profundo, al&#233;m das camadas de entrada e sa&#237;da, existem camadas ocultas que aumentam significativamente o n&#250;mero de pesos a serem atualizados e geralmente exigem grandes esfor&#231;os computacionais. A rede convolucional &#233; um tipo de aprendizado profundo inspirada no c&#243;rtex visual de animais que tem um papel importante na an&#225;lise de imagens. Autoencoders e redes neurais de Kohonen s&#227;o exemplos de aprendizado n&#227;o supervisionado. <xref ref-type="bibr" rid="B1"><sup>1</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B7"><sup>7</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B50"><sup>50</sup></xref><sup>-</sup><xref ref-type="bibr" rid="B52"><sup>52</sup></xref>
									</td>
                  <td align="center">N&#227;o supervisionado ou Supervisionado</td>
                </tr>
                <tr>
                  <td>GB</td>
                  <td align="center">&#201; um m&#233;todo baseado em &#225;rvore que utiliza gradientes, vetores relacionados &#224; dire&#231;&#227;o do aumento m&#225;ximo em uma fun&#231;&#227;o matem&#225;tica, para produzir &#225;rvores de decis&#227;o sequenciais a serem combinadas para aprimorar a previs&#227;o. Variantes desta abordagem incluem o Stochastic Gradient Descent que incorpora uma sub-amostra aleat&#243;ria para GB. <xref ref-type="bibr" rid="B53"><sup>53</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B54"><sup>54</sup></xref>
									</td>
                  <td align="center">Supervisionado</td>
                </tr>
              </tbody>
            </table>
          </table-wrap>
        </p>
        <list list-type="alpha-lower">
          <list-item>
            <p><bold><italic>Support Vector Machine</italic></bold><bold>(SVM):</bold> utilizada por Samad et al.,<sup><xref ref-type="bibr" rid="B22">22</xref></sup> para prever com sucesso a deteriora&#231;&#227;o da fun&#231;&#227;o ventricular em pacientes submetidos a reparo da tetralogia de Fallot a partir de um banco de dados de 153 pacientes com dados cl&#237;nicos, eletrocardiogr&#225;ficos e de resson&#226;ncia magn&#233;tica card&#237;aca. Em rela&#231;&#227;o &#224; previs&#227;o de qualquer deteriora&#231;&#227;o (menor ou maior) vs. nenhuma deteriora&#231;&#227;o, a m&#233;dia da &#225;rea sob a curva (AUC) foi de 0,82 &#177; 0,06.<sup><xref ref-type="bibr" rid="B22">22</xref></sup> Berikol et al.<sup><xref ref-type="bibr" rid="B23">23</xref></sup> utilizaram dados cl&#237;nicos, laboratoriais (n&#237;veis de troponina I e CK-MB), do ECG e ecocardiogr&#225;ficos de 228 pacientes que apresentaram dor no peito no pronto-socorro para classifica&#231;&#227;o quanto &#224; presen&#231;a ou aus&#234;ncia de S&#237;ndrome Coronariana Aguda. Precis&#227;o, sensibilidade e especificidade foram, respectivamente, 99,19, 98,22 e 100%.<sup><xref ref-type="bibr" rid="B23">23</xref></sup> Betancur et al.,<sup><xref ref-type="bibr" rid="B24">24</xref></sup> tamb&#233;m utilizaram o SVM para definir com maior precis&#227;o o posicionamento do plano da v&#225;lvula (PV) mitral durante a segmenta&#231;&#227;o ventricular esquerda nos exames de Tomografia Computadorizada por Emiss&#227;o de F&#243;ton &#218;nico (SPECT). Imagens de 392 pacientes foram analisadas e os bons resultados obtidos foram compat&#237;veis com a opini&#227;o de especialistas da &#225;rea &#8211; AUC: 0,82 [0,74-0,9] para detec&#231;&#227;o regional de &#225;reas de estenose obstrutiva e &#225;reas de d&#233;ficit de perfus&#227;o total isqu&#234;mica.<sup><xref ref-type="bibr" rid="B24">24</xref></sup></p>
          </list-item>
          <list-item>
            <p><bold><italic>Naive Bayes</italic></bold><bold>(NB):</bold> Paredes et al.,<sup><xref ref-type="bibr" rid="B25">25</xref></sup> utilizaram uma fus&#227;o de NB e algoritmo gen&#233;tico para prever o risco de ocorr&#234;ncia de eventos cardiovasculares (por exemplo, hospitaliza&#231;&#227;o ou morte), com base em dados de 559 pacientes com S&#237;ndrome Coronariana Aguda - Infarto do Mioc&#225;rdio sem Supradesnivelamento do Segmento ST (SCA-IAMSST). Sensibilidade e especificidade foram, respectivamente, 79,8, e 83,8.<sup><xref ref-type="bibr" rid="B25">25</xref></sup></p>
          </list-item>
          <list-item>
            <p><bold><italic>K-nearest neighbors</italic></bold><bold>(KNN):</bold> Al-Mallah et al.,<sup><xref ref-type="bibr" rid="B26">26</xref></sup> compararam a previs&#227;o de mortalidade por todas as causas em 10 anos entre o modelo de regress&#227;o log&#237;stica cl&#225;ssico e o KNN, considerando um banco de dados de 34.212 pacientes com informa&#231;&#245;es cl&#237;nicas e informa&#231;&#245;es obtidas ap&#243;s o teste de esfor&#231;o em esteira utilizando o protocolo padr&#227;o de Bruce.<sup><xref ref-type="bibr" rid="B26">26</xref></sup> Os resultados obtidos por essa ferramenta de ML mostraram sensibilidade de 87,4% e especificidade de 97,2%, melhores do que o desempenho preditivo do tradicional escore de risco <italic>Atherosclerosis Cardiovascular Disease Risk Score</italic> (ASCVD).</p>
          </list-item>
          <list-item>
            <p><bold><italic>Genetic algorithms</italic></bold><bold>(GA):</bold> Smisek et al.,<sup><xref ref-type="bibr" rid="B27">27</xref></sup> desenvolveram um dispositivo <italic>wearable</italic> (&#8220;vest&#237;vel&#8221;) para detectar arritmias a partir do registro de informa&#231;&#245;es de um eletrocardiograma de deriva&#231;&#227;o &#250;nica. Os dados foram analisados a partir de uma combina&#231;&#227;o do (SVM), &#225;rvore de decis&#227;o e regras baseadas em limiares. Algoritmos gen&#233;ticos foram usados para selecionar as caracter&#237;sticas mais adequadas a serem utilizadas no trabalho. Em rela&#231;&#227;o &#224; detec&#231;&#227;o de fibrila&#231;&#227;o atrial, obteve-se um escore F1 (m&#233;dia harm&#244;nica de valor preditivo positivo e sensibilidade) de 0,81.<sup><xref ref-type="bibr" rid="B27">27</xref></sup> Stuckey et al.,<sup><xref ref-type="bibr" rid="B28">28</xref></sup> utilizaram a An&#225;lise de Tomografia Espacial de Fase Card&#237;aca &#8211; um m&#233;todo pioneiro que dispensa o uso de radia&#231;&#227;o e contraste, bem como a realiza&#231;&#227;o de testes de esfor&#231;o ou farmacol&#243;gico &#8211; combinado com modelos de ML (por exemplo, algoritmos gen&#233;ticos) para analisar os sinais da fase tor&#225;cica. Neste estudo, os autores utilizaram essa ferramenta para avaliar pacientes com doen&#231;a coronariana e dor tor&#225;cica que foram encaminhados pelo m&#233;dico para realizar uma angiografia. Foram estudados 606 pacientes e os resultados mostraram sensibilidade de 92%, especificidade de 62% e valor preditivo de 96% para doen&#231;a coronariana.<sup><xref ref-type="bibr" rid="B28">28</xref></sup></p>
          </list-item>
          <list-item>
            <p><bold><italic>Random Forests</italic></bold><bold>(RF):</bold> Samad et al.,<sup><xref ref-type="bibr" rid="B29">29</xref></sup> analisaram um banco de dados composto por vari&#225;veis cl&#237;nicas e eletrocardiogr&#225;ficas para avaliar a sobrevida em 10 diferentes per&#237;odos de tempo (variando de 6 a 60 meses), considerando um total de 171.510 pacientes. A RF foi utilizada, com excelentes resultados, melhor do que aqueles obtidos por meio de escores tradicionais, como o escore de risco de Framingham e o escore das diretrizes da ACC/AHA. A &#225;rea sob a curva (AUC) foi superior a 0,82.<sup><xref ref-type="bibr" rid="B29">29</xref></sup> Ambale-Venkatesh, et al.,<sup><xref ref-type="bibr" rid="B30">30</xref></sup> utilizaram informa&#231;&#245;es de testes n&#227;o invasivos, question&#225;rios, biomarcadores e exames de imagem de 6.814 pacientes para construir 739 vari&#225;veis (caracter&#237;sticas), a fim de aplicar uma variante do RF &#8211; chamada <italic>survivor random forests</italic>
 <sup><xref ref-type="bibr" rid="B31">31</xref></sup> &#8211; para prever eventos cardiovasculares. (morte por todas as causas, acidente vascular cerebral, todas as doen&#231;as cardiovasculares, doen&#231;a coronariana, fibrila&#231;&#227;o atrial e insufici&#234;ncia card&#237;aca), que apresentou um desempenho melhor do que os escores de risco estabelecidos, como por exemplo, MESA-CHD, AHA/ASCVD e o escore de Framingham, com maior acur&#225;cia na previs&#227;o (diminui&#231;&#227;o de 10%-25% do escore de Brier)<sup><xref ref-type="bibr" rid="B30">30</xref> , <xref ref-type="bibr" rid="B31">31</xref></sup> .</p>
          </list-item>
          <list-item>
            <p><bold><italic>K-means</italic></bold><bold>:</bold> Cikes et al.,<sup><xref ref-type="bibr" rid="B32">32</xref></sup> utilizaram um banco de dados composto por vari&#225;veis cl&#237;nicas e par&#226;metros ecocardiogr&#225;ficos para os quais foram aplicados dois modelos de ML, <italic>K-means</italic> e <italic>Multiple Kernel Learning</italic> , a fim de categorizar os pacientes em grupos mutuamente exclusivos para avaliar a resposta &#224; terapia de ressincroniza&#231;&#227;o card&#237;aca. Foram analisados 1.106 pacientes e identificados quatro grupos disjuntos, dois deles com a melhor resposta &#224; terapia.<sup><xref ref-type="bibr" rid="B32">32</xref></sup></p>
          </list-item>
          <list-item>
            <p><bold><italic>Artificial Neural Networks</italic></bold><bold>(ANN):</bold> Kwon et al.,<sup><xref ref-type="bibr" rid="B33">33</xref></sup> em um estudo multic&#234;ntrico com 52.131 pacientes, constru&#237;ram um sistema de alerta precoce baseado em aprendizado profundo capaz de prever a ocorr&#234;ncia de parada card&#237;aca em um hospital. O modelo demonstrou alto desempenho quando comparado aos sistemas tradicionais do tipo &#8220; <italic>track-and-trigger</italic> &#8221; (rastreamento e disparo). A &#225;rea sob a curva foi de 0,82.<sup><xref ref-type="bibr" rid="B33">33</xref></sup> Rubin et al.,<sup><xref ref-type="bibr" rid="B34">34</xref></sup> tiveram resultados preliminares promissores com o uso de redes neurais com arquitetura convolucional para avaliar sinais eletrocardiogr&#225;ficos e classific&#225;-los em fibrila&#231;&#227;o atrial, ritmo sinusal (normal) ou ru&#237;do &#8211; o escore F1 alcan&#231;ado foi de 0,82.<sup><xref ref-type="bibr" rid="B34">34</xref></sup> Zhang et al.,<sup><xref ref-type="bibr" rid="B35">35</xref></sup> tamb&#233;m utilizaram redes neurais convolucionais para analisar um banco de dados com 14.035 exames ecocardiogr&#225;ficos para detectar a presen&#231;a de doen&#231;as como cardiomiopatia hipertr&#243;fica, amiloidose card&#237;aca e hipertens&#227;o arterial pulmonar com alto desempenho: as estat&#237;sticas C foram respectivamente de 0,93, 0,87 e 0,85.<sup><xref ref-type="bibr" rid="B35">35</xref></sup> Nakajima et al.,<sup><xref ref-type="bibr" rid="B36">36</xref></sup> utilizaram uma ANN para avaliar a presen&#231;a de doen&#231;a coronariana ap&#243;s a realiza&#231;&#227;o de cintilografia do mioc&#225;rdio. Os resultados foram obtidos com alta precis&#227;o e desempenho superior aos escores tradicionais utilizados. Por exemplo, a AUC para pacientes com infarto do mioc&#225;rdio antigo com base em defeitos na fase de repouso foi de 0,97.<sup><xref ref-type="bibr" rid="B36">36</xref></sup></p>
          </list-item>
          <list-item>
            <p><bold><italic>Gradient Boosting</italic></bold><bold>(GB):</bold> Mortazavi et al.,<sup><xref ref-type="bibr" rid="B37">37</xref></sup> utilizaram o GB para predizer o risco de sangramento ap&#243;s interven&#231;&#227;o coron&#225;ria percut&#226;nea e demonstraram que essas ferramentas podem ajudar a identificar pacientes que se beneficiariam de estrat&#233;gias objetivando a redu&#231;&#227;o do risco de sangramento. Foram analisados 3.316.465 procedimentos e obtida uma estat&#237;stica C de 0,82<sup><xref ref-type="bibr" rid="B37">37</xref></sup> . Hernesniemi et al.,<sup><xref ref-type="bibr" rid="B38">38</xref></sup> tamb&#233;m propuseram um GB para prever a mortalidade na s&#237;ndrome coronariana aguda, analisando 9.066 pacientes consecutivos. A AUC foi de 0,89 e o modelo apresentou melhor desempenho do que o do escore tradicional GRACE.<sup><xref ref-type="bibr" rid="B38">38</xref></sup></p>
          </list-item>
        </list>
        <p>&#201; importante observar que, ao utilizar qualquer modelo de ML, deve-se ter em mente um grande problema que pode surgir, chamado sobreajuste ( <italic>overfitting</italic> ). Ele ocorre quando um modelo descreve muito bem os exemplos (subconjunto de treinamento) e apresenta baixo desempenho quando aplicado a outras inst&#226;ncias do mesmo fen&#244;meno.<sup><xref ref-type="bibr" rid="B39">39</xref></sup> Al&#233;m disso, &#233; importante dizer que n&#227;o h&#225; resultado te&#243;rico que garanta que qualquer um dos algoritmos de IA seja melhor que os outros em qualquer aplica&#231;&#227;o.</p>
        <p>Portanto, essa escolha depende de diversas vari&#225;veis, como a natureza do problema em an&#225;lise, o tempo e os recursos dispon&#237;veis para resolv&#234;-lo. A combina&#231;&#227;o de t&#233;cnicas que geram modelos h&#237;bridos tamb&#233;m pode ser de grande valor. Por outro lado, o uso de ferramentas para processamento paralelo, como as GPUs ( <italic>Graphic Processing Units</italic> ), tem sido de grande valia para melhorar o desempenho dos modelos de ML, principalmente em rela&#231;&#227;o ao tempo computacional necess&#225;rio para execut&#225;-los.</p>
      </sec>
      <sec>
        <title>Desafios e perspectivas futuras</title>
        <p>Como destacado anteriormente, as aplica&#231;&#245;es de IA em cardiologia aumentaram bastante nos &#250;ltimos anos e seu potencial de crescimento &#233; enorme. No entanto, esse cen&#225;rio traz consigo a necessidade de superar alguns desafios, tais como: limites &#233;ticos de uso (uso indevido), aprimoramento do conhecimento matem&#225;tico, aquisi&#231;&#227;o de dados saud&#225;veis, desenvolvimento de seguran&#231;a, necessidade de colabora&#231;&#227;o, aten&#231;&#227;o a erros e cuidados baseados em dados. Tudo isso &#233; discutido abaixo e est&#225; resumido na <xref ref-type="fig" rid="f02">Figura 2</xref> .</p>
        <list list-type="alpha-lower">
          <list-item>
            <p><bold>Desafio 1 &#8211;</bold> limites &#233;ticos de uso (uso indevido): como toda tecnologia disruptiva, os limites da &#233;tica precisam ser repensados e amplamente discutidos. Os algoritmos de ML podem ser mal utilizados e enganosos. Como exemplo, uma obra de grande repercuss&#227;o foi publicada por Wang e Kosinski (2018). Os autores usaram aprendizado profundo e obtiveram resultados expressivos para prever se um indiv&#237;duo &#233; gay ou n&#227;o em um banco de dados de imagens dos rostos dos participantes do estudo.<sup><xref ref-type="bibr" rid="B55">55</xref></sup> De maneira similar, os mesmos algoritmos de IA podem ser utilizados para detectar, por exemplo, se um paciente desenvolver&#225; ou n&#227;o fibrila&#231;&#227;o atrial ou qualquer cardiomiopatia futura. Essas informa&#231;&#245;es poderiam ser utilizadas pelas empresas para aumentar os valores de seus planos de sa&#250;de ou at&#233; mesmo negar a associa&#231;&#227;o ao plano devido a um alto custo? E se for detectado que um beb&#234; nascer&#225; com doen&#231;a card&#237;aca cong&#234;nita devido &#224; an&#225;lise dos dados gen&#233;ticos, cl&#237;nico-laboratoriais e de imagem (ou outros) de seus pais? Isso poderia abrir espa&#231;o para um tipo de neoeugenia. Esse debate ganhou uma &#234;nfase adicional com o surgimento da t&#233;cnica CRISP-Cas9, que permite a edi&#231;&#227;o do DNA.<sup><xref ref-type="bibr" rid="B56">56</xref></sup> Nesse contexto, ao estimular um debate com a sociedade sobre o assunto, a transpar&#234;ncia e a regulamenta&#231;&#227;o s&#227;o pilares fundamentais a serem preservados.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 2 &#8211;</bold> melhorar o conhecimento de matem&#225;tica: o advento desse novo tipo de ser humano inacredit&#225;vel ( <italic>Homo incredibile</italic> ), que apoia suas decis&#245;es em dados, carrega consigo o papel fundamental da matem&#225;tica e da computa&#231;&#227;o nesta revolu&#231;&#227;o atualmente em andamento. Essa revolu&#231;&#227;o trar&#225; possibilidades inimagin&#225;veis na pr&#225;tica m&#233;dica, como a constru&#231;&#227;o de <italic>phenomappings</italic> de qualidade &#8211; modelos de ML desenvolvidos com o objetivo de agrupar pacientes em fun&#231;&#227;o de sua grande massa de caracter&#237;sticas fenot&#237;picas, a fim de facilitar o processo de tomada de decis&#227;o.<sup><xref ref-type="bibr" rid="B57">57</xref></sup> Assim, &#233; necess&#225;rio que essas compet&#234;ncias sejam estimuladas precocemente, principalmente com foco na solu&#231;&#227;o de problemas relacionados &#224; realidade para a qual se deseja promover melhorias. Isso certamente se refletir&#225; na necessidade de reformular o conte&#250;do cardiovascular (e por que n&#227;o dizer, o conte&#250;do m&#233;dico em geral) dos cursos de gradua&#231;&#227;o e p&#243;s-gradua&#231;&#227;o em Medicina: uma educa&#231;&#227;o passiva ou meramente expositiva, com uma carga extensa e priorizando a capacidade de mem&#243;ria do aluno, parece cada vez mais inadequada, pois se percebe que a Medicina deve ser um espa&#231;o de criatividade e gera&#231;&#227;o de valor.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 3 &#8211;</bold> obtenha dados saud&#225;veis: o uso de dados saud&#225;veis &#233; de valor fundamental para o sucesso dos algoritmos. Assim, &#233; necess&#225;rio que as unidades de sa&#250;de incentivem seus profissionais de sa&#250;de a respeito do rigor no n&#237;vel de preenchimento/obten&#231;&#227;o de dados, bem como a manuten&#231;&#227;o de qualquer fonte de dados, de formul&#225;rios, prontu&#225;rios eletr&#244;nicos, dados de imagem ou mesmo dados n&#227;o convencionais, como os obtidos por Medina et al.<sup><xref ref-type="bibr" rid="B58">58</xref></sup> &#8211; que desenvolveram uma ferramenta de sa&#250;de de redes sociais online ( <italic>Online Social Networks Health</italic> ) bem-sucedida, na qual o pr&#243;prio paciente insere anonimamente informa&#231;&#245;es de monitoramento de sa&#250;de, incluindo dados fisiol&#243;gicos, atividades di&#225;rias, estados emocionais e intera&#231;&#227;o com outros pacientes.<sup><xref ref-type="bibr" rid="B58">58</xref></sup> Portanto, o gerenciamento de dados se torna t&#227;o importante quanto outros comportamentos de rotina na medicina baseada em evid&#234;ncias, como lavagem adequada das m&#227;os ou at&#233; o uso de um desfibrilador durante uma parada card&#237;aca. Dessa maneira, a forma&#231;&#227;o de equipes de dados multidisciplinares e o treinamento constante das equipes assumem um papel primordial. Vale ressaltar que grande parte da lentid&#227;o e dificuldade que algumas unidades de sa&#250;de t&#234;m no uso de modelos de ML est&#225; ligada a dados saud&#225;veis ausentes ou incipientes.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 4 &#8211;</bold> obtenha seguran&#231;a: o advento dessas ferramentas traz consigo uma preocupa&#231;&#227;o fundamental com a seguran&#231;a dos dados, a um n&#237;vel nunca experimentado, pois o acesso a esses dados por pessoas n&#227;o autorizadas pode levar a consequ&#234;ncias catastr&#243;ficas para as institui&#231;&#245;es de sa&#250;de e para os pacientes. A cria&#231;&#227;o de uma equipe de seguran&#231;a desempenha um papel importante nesse novo processo. O Regulamento Geral de Prote&#231;&#227;o de Dados ( <italic>General Data Protection Regulation</italic> ) representa um avan&#231;o nessa dire&#231;&#227;o. <italic>Blockchain</italic> e suas variantes s&#227;o ferramentas importantes que podem melhorar substancialmente a seguran&#231;a.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 5 &#8211;</bold> necessidade de colabora&#231;&#227;o (o prov&#233;rbio Africano): h&#225; um prov&#233;rbio africano que diz: &#8220; <italic>se voc&#234; quer ir r&#225;pido, v&#225; sozinho, mas se quiser ir longe, v&#225; com muitos</italic> &#8221;. Isso se aplica muito a esse ambiente de dados: a colabora&#231;&#227;o entre institui&#231;&#245;es permite a constru&#231;&#227;o de enormes bancos de dados saud&#225;veis ( <italic>Big Data</italic> ), o que tende a favorecer o desempenho dos algoritmos de ML.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 6 &#8211;</bold> lidando com erros: Uma quest&#227;o importante diz respeito aos erros dos modelos de IA. &#201; inadequado acreditar que esses modelos est&#227;o livres de erros. Por exemplo, pode ser o resultado de sobreajuste ou da ocorr&#234;ncia de dados n&#227;o &#237;ntegros &#8211; que tornam os resultados n&#227;o confi&#225;veis. No entanto, a pr&#225;tica mostrou alto desempenho em v&#225;rias aplica&#231;&#245;es. Esses modelos s&#227;o probabil&#237;sticos e &#233; sempre desej&#225;vel que seus erros sejam m&#237;nimos. Esse cen&#225;rio tem implica&#231;&#245;es cl&#237;nicas, por exemplo, um modelo de IA que prev&#234; com 99% de probabilidade que um paciente tem maior propens&#227;o do que a popula&#231;&#227;o em geral a ter miocardite card&#237;aca ou amiloidose. H&#225; uma probabilidade, embora pequena, de que isso n&#227;o ocorra e de que o procedimento adotado pelo cardiologista seja inadequado. Nesse caso, a quest&#227;o &#233; quem pode ser responsabilizado nesses casos? &#201; apropriado? O paciente deve assinar um termo de consentimento nesses casos? Certamente, a solu&#231;&#227;o inclui uma regulamenta&#231;&#227;o robusta do uso dessas ferramentas e o fortalecimento de um novo tipo de relacionamento: m&#233;dico-paciente-dados.</p>
          </list-item>
          <list-item>
            <p><bold>Desafio 7 &#8211;</bold> gest&#227;o de cuidados baseados em dados: enquanto as ferramentas de ML seguem um caminho inexor&#225;vel, por outro lado, v&#225;rios profissionais de sa&#250;de continuam com medo dessas ferramentas devido a sua poss&#237;vel capacidade de substituir os m&#233;dicos em suas tarefas. Entretanto, quando a hist&#243;ria da Medicina &#233; lembrada, vale ressaltar, por exemplo, que o aparecimento de m&#225;quinas automatizadas para realizar o hemograma n&#227;o substituiu o hematologista, mas resultou em uma maior velocidade do processo de trabalho e permitiu ao profissional atuar em outras quest&#245;es importantes da especialidade. A ideia central &#233; fornecer melhor suporte para a tomada de decis&#227;o, incluindo melhor desempenho. Trata-se da gest&#227;o assistencial baseada em dados, com alto dinamismo e atualiza&#231;&#227;o constante &#8211; o que promover&#225; maior personaliza&#231;&#227;o do cuidado<sup><xref ref-type="bibr" rid="B59">59</xref></sup> e avalia&#231;&#227;o em tempo real da experi&#234;ncia dos usu&#225;rios do sistema de sa&#250;de, visando gerar valor para o paciente. Nesse contexto, as tarefas mec&#226;nicas ser&#227;o substitu&#237;veis e uma diversidade de novas tarefas ser&#225; inclu&#237;da na rotina do cardiologista de precis&#227;o, desde a constru&#231;&#227;o adequada das bases de dados at&#233; a reflex&#227;o cr&#237;tica sobre os resultados obtidos pelos modelos matem&#225;tico-computacionais, bem como o desenvolvimento de um relacionamento m&#233;dico-paciente-dados adequado. Portanto, h&#225; uma migra&#231;&#227;o de habilidades humanas, bem como a expans&#227;o de suas capacidades a partir do surgimento de novas ferramentas, que devem fazer parte do arsenal t&#233;cnico do cardiologista do s&#233;culo XXI. Esse panorama nos permite comparar os modelos de ML a um cavalo e os m&#233;dicos a j&#243;queis: &#8220; <italic>quem corre &#233; o cavalo, voc&#234; precisa ser o j&#243;quei</italic> &#8221;.</p>
          </list-item>
        </list>
      </sec>
    </sec>
    <sec sec-type="conclusions">
      <title>Conclus&#245;es</title>
      <p>A IA, de fato, tem se mostrado uma ferramenta fundamental para a pr&#225;tica cl&#237;nica da cardiologia atual. Diversas aplica&#231;&#245;es foram realizadas com sucesso e permitiram melhorias significativas do ponto de vista diagn&#243;stico e terap&#234;utico e em rela&#231;&#227;o ao atendimento personalizado. Para poder utilizar tais ferramentas, &#233; imperativo que dados saud&#225;veis sejam utilizados, o que certamente implica um novo design no <italic>modus operandi</italic> de muitos servi&#231;os de sa&#250;de. A natureza desses dados &#233; variada e inclui novas fontes, como dispositivos &#8220;vest&#237;veis&#8221; e dados <italic>omic</italic> . Por outro lado, esse novo ecossistema digital requer uma aquisi&#231;&#227;o de conhecimento que n&#227;o &#233; tradicionalmente encontrado em cursos regulares de medicina. Portanto, um redesenho curricular &#233; necess&#225;rio e deve ser objeto de um profundo debate e a&#231;&#245;es espec&#237;ficas.</p>
      <p>Por outro lado, toda a panaceia trazida pela IA n&#227;o est&#225; isenta de desafios, como os limites &#233;ticos de seu uso, a necessidade de aprimorar o conhecimento matem&#225;tico, a constru&#231;&#227;o de um ecossistema que garanta altos n&#237;veis de seguran&#231;a e confidencialidade para os pacientes, a aquisi&#231;&#227;o de dados saud&#225;veis, as necessidades de expandir a associa&#231;&#227;o m&#233;dico-paciente-dados, a necessidade de colabora&#231;&#227;o e o gerenciamento de cuidados baseados em dados. Nesse contexto, o cardiologista-j&#243;quei (ou m&#233;dicos em geral) deve ser protagonista das mudan&#231;as e deve substituir um eventual medo das ferramentas por um maior envolvimento, com o objetivo de gerar valor para o cuidado. &#201; importante ter em mente os poss&#237;veis desafios e obst&#225;culos a serem superados e manter um engajamento e senso cr&#237;tico na busca de solu&#231;&#245;es: &#8220;quem corre &#233; o cavalo, voc&#234; precisa ser o j&#243;quei&#8221;.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <title>Refer&#234;ncias</title>
      <ref id="B1">
        <label>1</label>
        <mixed-citation>1. Goodfellow I, Bengio Y, Courville A. Deep Learning. Cambridge: MIT Press; 2016.</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Goodfellow</surname>
              <given-names>I</given-names>
            </name>
            <name>
              <surname>Bengio</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Courville</surname>
              <given-names>A</given-names>
            </name>
          </person-group>
          <source>Deep Learning</source>
          <publisher-loc>Cambridge</publisher-loc>
          <publisher-name>MIT Press</publisher-name>
          <year>2016</year>
        </element-citation>
      </ref>
      <ref id="B2">
        <label>2</label>
        <mixed-citation>2. Dilsizian SE, Siegel, EL. Artificial intelligence in medicine and cardiac imaging: harnessing big data and advanced computing to provide personalized medical diagnosis and treatment. Curr Cardiol Rep. 2014;16(1):441.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Dilsizian</surname>
              <given-names>SE</given-names>
            </name>
            <name>
              <surname>Siegel</surname>
              <given-names>EL</given-names>
            </name>
          </person-group>
          <article-title>Artificial intelligence in medicine and cardiac imaging: harnessing big data and advanced computing to provide personalized medical diagnosis and treatment</article-title>
          <source>Curr Cardiol Rep</source>
          <year>2014</year>
          <volume>16</volume>
          <issue>1</issue>
          <size units="pages">441</size>
        </element-citation>
      </ref>
      <ref id="B3">
        <label>3</label>
        <mixed-citation>3. Michel JB, Sangha DM, Erwin JP 3rd. Burnout among cardiologists. Am J Cardiol. 2017;119(6):938-40.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Michel</surname>
              <given-names>JB</given-names>
            </name>
            <name>
              <surname>Sangha</surname>
              <given-names>DM</given-names>
            </name>
            <name>
              <surname>Erwin</surname>
              <given-names>JP</given-names>
              <suffix>3rd</suffix>
            </name>
          </person-group>
          <article-title>Burnout among cardiologists</article-title>
          <source>Am J Cardiol</source>
          <year>2017</year>
          <volume>119</volume>
          <issue>6</issue>
          <fpage>938</fpage>
          <lpage>940</lpage>
        </element-citation>
      </ref>
      <ref id="B4">
        <label>4</label>
        <mixed-citation>4. Rotenstein LS, Torre M, Ramos MA, Rosales RC, Guille C, Sen S, et al. Prevalence of burnout among physicians: a systematic review. JAMA. 2018;320(11):1131-50.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Rotenstein</surname>
              <given-names>LS</given-names>
            </name>
            <name>
              <surname>Torre</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Ramos</surname>
              <given-names>MA</given-names>
            </name>
            <name>
              <surname>Rosales</surname>
              <given-names>RC</given-names>
            </name>
            <name>
              <surname>Guille</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Sen</surname>
              <given-names>S</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Prevalence of burnout among physicians: a systematic review</article-title>
          <source>JAMA</source>
          <year>2018</year>
          <volume>320</volume>
          <issue>11</issue>
          <fpage>1131</fpage>
          <lpage>1150</lpage>
        </element-citation>
      </ref>
      <ref id="B5">
        <label>5</label>
        <mixed-citation>5. Macary MA, Daniel M. Medical error &#8211; the third leading cause of death in the US. BMJ. 2016 May 3;353:i2139.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Macary</surname>
              <given-names>MA</given-names>
            </name>
            <name>
              <surname>Daniel</surname>
              <given-names>M</given-names>
            </name>
          </person-group>
          <article-title>Medical error &#8211; the third leading cause of death in the US</article-title>
          <source>BMJ</source>
          <year>2016</year>
          <month>05</month>
          <day>3</day>
          <volume>353</volume>
          <size units="pages">i2139</size>
        </element-citation>
      </ref>
      <ref id="B6">
        <label>6</label>
        <mixed-citation>6. Porter ME. A strategy for health care reform &#8211; toward a value-based system. N Engl J Med. 2009;361(2):109-12.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Porter</surname>
              <given-names>ME</given-names>
            </name>
          </person-group>
          <article-title>A strategy for health care reform &#8211; toward a value-based system</article-title>
          <source>N Engl J Med</source>
          <year>2009</year>
          <volume>361</volume>
          <issue>2</issue>
          <fpage>109</fpage>
          <lpage>112</lpage>
        </element-citation>
      </ref>
      <ref id="B7">
        <label>7</label>
        <mixed-citation>7. Dilsizian ME, Siegel EL. Machine meets biology: a primer on artificial intelligence in cardiology and cardiac imaging. Curr Cardiol Rep. 2018;20(12):139.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Dilsizian</surname>
              <given-names>ME</given-names>
            </name>
            <name>
              <surname>Siegel</surname>
              <given-names>EL</given-names>
            </name>
          </person-group>
          <article-title>Machine meets biology: a primer on artificial intelligence in cardiology and cardiac imaging</article-title>
          <source>Curr Cardiol Rep</source>
          <year>2018</year>
          <volume>20</volume>
          <issue>12</issue>
          <size units="pages">139</size>
        </element-citation>
      </ref>
      <ref id="B8">
        <label>8</label>
        <mixed-citation>8. Johnson KW, Soto JT, Glicksberg BS, Shameer K, Miotto R, Ali M, et al. Artificial Intelligence in Cardiology. J Am Coll Cardiol. 2018;71(23):2668-79.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Johnson</surname>
              <given-names>KW</given-names>
            </name>
            <name>
              <surname>Soto</surname>
              <given-names>JT</given-names>
            </name>
            <name>
              <surname>Glicksberg</surname>
              <given-names>BS</given-names>
            </name>
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Miotto</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Ali</surname>
              <given-names>M</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Artificial Intelligence in Cardiology</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2018</year>
          <volume>71</volume>
          <issue>23</issue>
          <fpage>2668</fpage>
          <lpage>2679</lpage>
        </element-citation>
      </ref>
      <ref id="B9">
        <label>9</label>
        <mixed-citation>9. Krittanawong C, Zhang H, Wang Z, Aydar M, Kitai T. Artificial intelligence in precision cardiovascular medicine. J Am Coll Cardiol. 2017;69(21):2657-64.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Krittanawong</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>Z</given-names>
            </name>
            <name>
              <surname>Aydar</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Kitai</surname>
              <given-names>T</given-names>
            </name>
          </person-group>
          <article-title>Artificial intelligence in precision cardiovascular medicine</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2017</year>
          <volume>69</volume>
          <issue>21</issue>
          <fpage>2657</fpage>
          <lpage>2664</lpage>
        </element-citation>
      </ref>
      <ref id="B10">
        <label>10</label>
        <mixed-citation>10. Massalha S, Clarkin O, Thornhill R, Wells G, Chow BJW. Decision support tools, systems, and artificial intelligence in cardiac imaging. Can J Cardiol. 2018;34(7):827-38.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Massalha</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Clarkin</surname>
              <given-names>O</given-names>
            </name>
            <name>
              <surname>Thornhill</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Wells</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Chow</surname>
              <given-names>BJW</given-names>
            </name>
          </person-group>
          <article-title>Decision support tools, systems, and artificial intelligence in cardiac imaging</article-title>
          <source>Can J Cardiol</source>
          <year>2018</year>
          <volume>34</volume>
          <issue>7</issue>
          <fpage>827</fpage>
          <lpage>838</lpage>
        </element-citation>
      </ref>
      <ref id="B11">
        <label>11</label>
        <mixed-citation>11. Mesquita CT. Artificial intelligence and machine learning in cardiology &#8211; a change of paradigm. Int. J. Cardiovasc. Sci. 2017;30(3):187-8.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Mesquita</surname>
              <given-names>CT</given-names>
            </name>
          </person-group>
          <article-title>Artificial intelligence and machine learning in cardiology &#8211; a change of paradigm</article-title>
          <source>Int. J. Cardiovasc. Sci</source>
          <year>2017</year>
          <volume>30</volume>
          <issue>3</issue>
          <fpage>187</fpage>
          <lpage>188</lpage>
        </element-citation>
      </ref>
      <ref id="B12">
        <label>12</label>
        <mixed-citation>12. Moore J. The Dartmouth College Artificial Intelligence Conference: The Next Fifty Years. AI Magazine. 2006;27(4):87-91.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Moore</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <article-title>The Dartmouth College Artificial Intelligence Conference: The Next Fifty Years</article-title>
          <source>AI Magazine</source>
          <year>2006</year>
          <volume>27</volume>
          <issue>4</issue>
          <fpage>87</fpage>
          <lpage>91</lpage>
        </element-citation>
      </ref>
      <ref id="B13">
        <label>13</label>
        <mixed-citation>13. Turing AM. Computing machinery and intelligence. Mind. 1950;59(236):433-60.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Turing</surname>
              <given-names>AM</given-names>
            </name>
          </person-group>
          <article-title>Computing machinery and intelligence</article-title>
          <source>Mind</source>
          <year>1950</year>
          <volume>59</volume>
          <issue>236</issue>
          <fpage>433</fpage>
          <lpage>460</lpage>
        </element-citation>
      </ref>
      <ref id="B14">
        <label>14</label>
        <mixed-citation>14. Price, WN. Big data and black-box medical algorithms. Sci Transl Med. 2018;10(471):pi:eaao5333.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Price</surname>
              <given-names>WN</given-names>
            </name>
          </person-group>
          <article-title>Big data and black-box medical algorithms</article-title>
          <source>Sci Transl Med</source>
          <year>2018</year>
          <volume>10</volume>
          <issue>471</issue>
          <comment>pi:eaao5333</comment>
        </element-citation>
      </ref>
      <ref id="B15">
        <label>15</label>
        <mixed-citation>15. Antman EM, Loscalzo J. Precision medicine in cardiology. Nat Rev Cardiol. 2016;13(10):591-602.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Antman</surname>
              <given-names>EM</given-names>
            </name>
            <name>
              <surname>Loscalzo</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <article-title>Precision medicine in cardiology</article-title>
          <source>Nat Rev Cardiol</source>
          <year>2016</year>
          <volume>13</volume>
          <issue>10</issue>
          <fpage>591</fpage>
          <lpage>602</lpage>
        </element-citation>
      </ref>
      <ref id="B16">
        <label>16</label>
        <mixed-citation>16. Price, WN II. Black Box medicine. Harvard J Law Tech. 2015;28(2):419-467.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Price</surname>
              <given-names>WN II</given-names>
            </name>
          </person-group>
          <article-title>Black Box medicine</article-title>
          <source>Harvard J Law Tech</source>
          <year>2015</year>
          <volume>28</volume>
          <issue>2</issue>
          <fpage>419</fpage>
          <lpage>467</lpage>
        </element-citation>
      </ref>
      <ref id="B17">
        <label>17</label>
        <mixed-citation>17. Dey D, Slomka PJ, Leeson P, Comaniciu D, Shrestha S, Sengupta PP, et al. Artificial intelligence in cardiovascular imaging: JACC State-of-the-Art review. J Am Coll Cardiol. 2019;73(11):1317-35.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Dey</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Slomka</surname>
              <given-names>PJ</given-names>
            </name>
            <name>
              <surname>Leeson</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Comaniciu</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Shrestha</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Sengupta</surname>
              <given-names>PP</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Artificial intelligence in cardiovascular imaging: JACC State-of-the-Art review</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2019</year>
          <volume>73</volume>
          <issue>11</issue>
          <fpage>1317</fpage>
          <lpage>1335</lpage>
        </element-citation>
      </ref>
      <ref id="B18">
        <label>18</label>
        <mixed-citation>18. Shameer K, Johnson KW, Glicksberg BS, Dudley JT, Sengupta PP. Machine learning in cardiovascular medicine: are we there yet? Heart. 2018;104(14):1156-64.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Johnson</surname>
              <given-names>KW</given-names>
            </name>
            <name>
              <surname>Glicksberg</surname>
              <given-names>BS</given-names>
            </name>
            <name>
              <surname>Dudley</surname>
              <given-names>JT</given-names>
            </name>
            <name>
              <surname>Sengupta</surname>
              <given-names>PP</given-names>
            </name>
          </person-group>
          <article-title>Machine learning in cardiovascular medicine: are we there yet?</article-title>
          <source>Heart</source>
          <year>2018</year>
          <volume>104</volume>
          <issue>14</issue>
          <fpage>1156</fpage>
          <lpage>1164</lpage>
        </element-citation>
      </ref>
      <ref id="B19">
        <label>19</label>
        <mixed-citation>19. Al&#8217;Aref SJ, Anchouche K, Singh G, Slomka PJ, Kolli KK, Kumar A, et al. Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging. Eur Heart J. 2018;40(24):1975-86.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Al&#8217;Aref</surname>
              <given-names>SJ</given-names>
            </name>
            <name>
              <surname>Anchouche</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Singh</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Slomka</surname>
              <given-names>PJ</given-names>
            </name>
            <name>
              <surname>Kolli</surname>
              <given-names>KK</given-names>
            </name>
            <name>
              <surname>Kumar</surname>
              <given-names>A</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging</article-title>
          <source>Eur Heart J</source>
          <year>2018</year>
          <volume>40</volume>
          <issue>24</issue>
          <fpage>1975</fpage>
          <lpage>1986</lpage>
        </element-citation>
      </ref>
      <ref id="B20">
        <label>20</label>
        <mixed-citation>20. Somashekhar SP, Sepulveda MJ, Puglielli S, Norden AD, Shortliffe EH, Rohit Kumar C, et al. Watson for Oncology and breast cancer treatment recommendations: agreement with an expert multidisciplinary tumor board. Ann Oncol. 2018;29(2):418-23.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Somashekhar</surname>
              <given-names>SP</given-names>
            </name>
            <name>
              <surname>Sepulveda</surname>
              <given-names>MJ</given-names>
            </name>
            <name>
              <surname>Puglielli</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Norden</surname>
              <given-names>AD</given-names>
            </name>
            <name>
              <surname>Shortliffe</surname>
              <given-names>EH</given-names>
            </name>
            <name>
              <surname>Rohit Kumar</surname>
              <given-names>C</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Watson for Oncology and breast cancer treatment recommendations: agreement with an expert multidisciplinary tumor board</article-title>
          <source>Ann Oncol</source>
          <year>2018</year>
          <volume>29</volume>
          <issue>2</issue>
          <fpage>418</fpage>
          <lpage>423</lpage>
        </element-citation>
      </ref>
      <ref id="B21">
        <label>21</label>
        <mixed-citation>21. IBM. Watson. New York: IBM; 2019 [Cited in 2018 Oct 10]. Available from: <ext-link ext-link-type="uri" xlink:href="https://www.ibm.com/watson/index.html">https://www.ibm.com/watson/index.html</ext-link>.</mixed-citation>
        <element-citation publication-type="webpage">
          <person-group person-group-type="author">
            <collab>IBM</collab>
          </person-group>
          <source>Watson</source>
          <publisher-loc>New York</publisher-loc>
          <publisher-name>IBM</publisher-name>
          <year>2019</year>
          <date-in-citation content-type="cited-date">Cited in 2018 Oct 10</date-in-citation>
          <comment>
            <ext-link ext-link-type="uri" xlink:href="https://www.ibm.com/watson/index.html">https://www.ibm.com/watson/index.html</ext-link>
          </comment>
        </element-citation>
      </ref>
      <ref id="B22">
        <label>22</label>
        <mixed-citation>22. Samad MD, Wehner GJ, Arbabshirani MR, Jing L, Powell AJ, Geva T, et al. Predicting deterioration of ventricular function in patients with repaired tetralogy of Fallot using machine learning. Eur Heart J Cardiovasc Imaging. 2018;19(7):730-8.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Samad</surname>
              <given-names>MD</given-names>
            </name>
            <name>
              <surname>Wehner</surname>
              <given-names>GJ</given-names>
            </name>
            <name>
              <surname>Arbabshirani</surname>
              <given-names>MR</given-names>
            </name>
            <name>
              <surname>Jing</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Powell</surname>
              <given-names>AJ</given-names>
            </name>
            <name>
              <surname>Geva</surname>
              <given-names>T</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Predicting deterioration of ventricular function in patients with repaired tetralogy of Fallot using machine learning</article-title>
          <source>Eur Heart J Cardiovasc Imaging</source>
          <year>2018</year>
          <volume>19</volume>
          <issue>7</issue>
          <fpage>730</fpage>
          <lpage>738</lpage>
        </element-citation>
      </ref>
      <ref id="B23">
        <label>23</label>
        <mixed-citation>23. Berikol GB, Yildiz O, &#214;zcan IT. Diagnosis of acute coronary syndrome with a support vector machine. J Med Syst. 2016;40(4):84.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Berikol</surname>
              <given-names>GB</given-names>
            </name>
            <name>
              <surname>Yildiz</surname>
              <given-names>O</given-names>
            </name>
            <name>
              <surname>&#214;zcan</surname>
              <given-names>IT</given-names>
            </name>
          </person-group>
          <article-title>Diagnosis of acute coronary syndrome with a support vector machine</article-title>
          <source>J Med Syst</source>
          <year>2016</year>
          <volume>40</volume>
          <issue>4</issue>
          <size units="pages">84</size>
        </element-citation>
      </ref>
      <ref id="B24">
        <label>24</label>
        <mixed-citation>24. Betancur J, Rubeaux M, Fuchs TA, Otaki Y, Arnson Y, Slipczuk L, et al. Automatic valve plane localization in myocardial perfusion SPECT/CT by machine learning: Anatomic and clinical validation. J Nucl Med. 2016;58(6):961-7.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Betancur</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Rubeaux</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Fuchs</surname>
              <given-names>TA</given-names>
            </name>
            <name>
              <surname>Otaki</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Arnson</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Slipczuk</surname>
              <given-names>L</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Automatic valve plane localization in myocardial perfusion SPECT/CT by machine learning: Anatomic and clinical validation</article-title>
          <source>J Nucl Med</source>
          <year>2016</year>
          <volume>58</volume>
          <issue>6</issue>
          <fpage>961</fpage>
          <lpage>967</lpage>
        </element-citation>
      </ref>
      <ref id="B25">
        <label>25</label>
        <mixed-citation>25. Paredes S, Rocha T, de Carvalho P, Henriques J, Morais J, Ferreira J. Integration of different risk assessment tools to improve stratification of patients with coronary artery disease. Med Biol Eng Comput. 2015;53(10):1069-83.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Paredes</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Rocha</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Carvalho</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Henriques</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Morais</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Ferreira</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <article-title>Integration of different risk assessment tools to improve stratification of patients with coronary artery disease</article-title>
          <source>Med Biol Eng Comput</source>
          <year>2015</year>
          <volume>53</volume>
          <issue>10</issue>
          <fpage>1069</fpage>
          <lpage>1083</lpage>
        </element-citation>
      </ref>
      <ref id="B26">
        <label>26</label>
        <mixed-citation>26. Al-Mallah MH, Elshawi R, Ahmed AM, Qureshi WT, Brawner CA, Blaha MJ, et al. Using machine learning to the association between cardiorespiratory fitness and all-cause mortality (from the Henry Ford Exercise Testing Project). Am J Cardiol. 2017;120(11):2078-84.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Al-Mallah</surname>
              <given-names>MH</given-names>
            </name>
            <name>
              <surname>Elshawi</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Ahmed</surname>
              <given-names>AM</given-names>
            </name>
            <name>
              <surname>Qureshi</surname>
              <given-names>WT</given-names>
            </name>
            <name>
              <surname>Brawner</surname>
              <given-names>CA</given-names>
            </name>
            <name>
              <surname>Blaha</surname>
              <given-names>MJ</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Using machine learning to the association between cardiorespiratory fitness and all-cause mortality (from the Henry Ford Exercise Testing Project)</article-title>
          <source>Am J Cardiol</source>
          <year>2017</year>
          <volume>120</volume>
          <issue>11</issue>
          <fpage>2078</fpage>
          <lpage>2084</lpage>
        </element-citation>
      </ref>
      <ref id="B27">
        <label>27</label>
        <mixed-citation>27. Smisek R, Hejc J, Ronzhina M, Nemcova A, Marsanova L, Kolarova J, et al. Multi-stage SVM approach for cardiac arrhythmias detection in short single-lead ECG recorded by a wearable device. Physiol Meas. 2018;39(9):094003.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Smisek</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Hejc</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Ronzhina</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Nemcova</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Marsanova</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Kolarova</surname>
              <given-names>J</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Multi-stage SVM approach for cardiac arrhythmias detection in short single-lead ECG recorded by a wearable device</article-title>
          <source>Physiol Meas</source>
          <year>2018</year>
          <volume>39</volume>
          <issue>9</issue>
          <size units="pages">094003</size>
        </element-citation>
      </ref>
      <ref id="B28">
        <label>28</label>
        <mixed-citation>28. Stuckey TD, Gammon RS, Goswami R, Depta JP, Steuter JA, Meine FJ 3rd, et al. Cardiac phase space tomography: a novel method of assessing coronary artery disease utilizing machine learning. PLoS One. 2018;13(8):e0198603.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Stuckey</surname>
              <given-names>TD</given-names>
            </name>
            <name>
              <surname>Gammon</surname>
              <given-names>RS</given-names>
            </name>
            <name>
              <surname>Goswami</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Depta</surname>
              <given-names>JP</given-names>
            </name>
            <name>
              <surname>Steuter</surname>
              <given-names>JA</given-names>
            </name>
            <name>
              <surname>Meine</surname>
              <given-names>FJ</given-names>
              <suffix>3rd</suffix>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Cardiac phase space tomography: a novel method of assessing coronary artery disease utilizing machine learning</article-title>
          <source>PLoS One</source>
          <year>2018</year>
          <volume>13</volume>
          <issue>8</issue>
          <elocation-id>e0198603</elocation-id>
        </element-citation>
      </ref>
      <ref id="B29">
        <label>29</label>
        <mixed-citation>29. Samad MD, Ulloa A, Wehner GJ, Jing L, Hartzel D, Good CW, et al. Predicting survival from large echocardiography and electronic health record datasets: optimization with machine learning. JACC Cardiovasc Imaging. 2019;12(4):681-9.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Samad</surname>
              <given-names>MD</given-names>
            </name>
            <name>
              <surname>Ulloa</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Wehner</surname>
              <given-names>GJ</given-names>
            </name>
            <name>
              <surname>Jing</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Hartzel</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Good</surname>
              <given-names>CW</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Predicting survival from large echocardiography and electronic health record datasets: optimization with machine learning</article-title>
          <source>JACC Cardiovasc Imaging</source>
          <year>2019</year>
          <volume>12</volume>
          <issue>4</issue>
          <fpage>681</fpage>
          <lpage>689</lpage>
        </element-citation>
      </ref>
      <ref id="B30">
        <label>30</label>
        <mixed-citation>30. Ambale-Venkatesh B, Yang X, Wu CO, Liu K, Hundley WG, McClelland R, et al. Cardiovascular event prediction by machine learning: the multi-ethnic study of atherosclerosis. Circ Res.2017;121(9):1092-101.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ambale-Venkatesh</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Yang</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>CO</given-names>
            </name>
            <name>
              <surname>Liu</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Hundley</surname>
              <given-names>WG</given-names>
            </name>
            <name>
              <surname>McClelland</surname>
              <given-names>R</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Cardiovascular event prediction by machine learning: the multi-ethnic study of atherosclerosis</article-title>
          <source>Circ Res</source>
          <year>2017</year>
          <volume>121</volume>
          <issue>9</issue>
          <fpage>1092</fpage>
          <lpage>1101</lpage>
        </element-citation>
      </ref>
      <ref id="B31">
        <label>31</label>
        <mixed-citation>31. Ishwaran H, Kogalur UB, Blackstone EH, Lauer MS. Random survival forests. Ann Appl Stat. 2008;2(3):841-60.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ishwaran</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Kogalur</surname>
              <given-names>UB</given-names>
            </name>
            <name>
              <surname>Blackstone</surname>
              <given-names>EH</given-names>
            </name>
            <name>
              <surname>Lauer</surname>
              <given-names>MS</given-names>
            </name>
          </person-group>
          <article-title>Random survival forests</article-title>
          <source>Ann Appl Stat</source>
          <year>2008</year>
          <volume>2</volume>
          <issue>3</issue>
          <fpage>841</fpage>
          <lpage>860</lpage>
        </element-citation>
      </ref>
      <ref id="B32">
        <label>32</label>
        <mixed-citation>32. Cikes M, Sanchez-Martinez S, Claggett B, Duchateau N, Piella G, Butakoff C, et al. Machine learning-based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy. Eur J Heart Fail. 2019;21(1):74-85.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Cikes</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Sanchez-Martinez</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Claggett</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Duchateau</surname>
              <given-names>N</given-names>
            </name>
            <name>
              <surname>Piella</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Butakoff</surname>
              <given-names>C</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Machine learning-based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy</article-title>
          <source>Eur J Heart Fail</source>
          <year>2019</year>
          <volume>21</volume>
          <issue>1</issue>
          <fpage>74</fpage>
          <lpage>85</lpage>
        </element-citation>
      </ref>
      <ref id="B33">
        <label>33</label>
        <mixed-citation>33. Kwon JM, Lee Y, Lee Y, Lee S, Park J 2018;7(13):pii: e008678.</mixed-citation>
        <element-citation publication-type="other">
          <person-group person-group-type="author">
            <name>
              <surname>Kwon</surname>
              <given-names>JM</given-names>
            </name>
            <name>
              <surname>Lee</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Lee</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Lee</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Park</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <year>2018</year>
          <volume>7</volume>
          <issue>13</issue>
          <comment>pii: e008678</comment>
        </element-citation>
      </ref>
      <ref id="B34">
        <label>34</label>
        <mixed-citation>34. Rubin J, Parvaneh S, Rahman A, Conroy B, Babaeizadeh S. Densely connected convolutional networks for detection of atrial fibrillation from short single-lead ECG recordings. J Electrocardiol. 2018;51(6S):S18-S21.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Rubin</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Parvaneh</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Rahman</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Conroy</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Babaeizadeh</surname>
              <given-names>S</given-names>
            </name>
          </person-group>
          <article-title>Densely connected convolutional networks for detection of atrial fibrillation from short single-lead ECG recordings</article-title>
          <source>J Electrocardiol</source>
          <year>2018</year>
          <volume>51</volume>
          <issue>6S</issue>
          <fpage>S18</fpage>
          <lpage>S21</lpage>
        </element-citation>
      </ref>
      <ref id="B35">
        <label>35</label>
        <mixed-citation>35. Zhang J, Gajjala S, Agrawal P, Tison GH, Hallock LA, Beussink-Nelson L, et al. Fully automated echocardiogram interpretation in clinical practice. Circulation. 2018;138(16):1623-35.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Zhang</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Gajjala</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Agrawal</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Tison</surname>
              <given-names>GH</given-names>
            </name>
            <name>
              <surname>Hallock</surname>
              <given-names>LA</given-names>
            </name>
            <name>
              <surname>Beussink-Nelson</surname>
              <given-names>L</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Fully automated echocardiogram interpretation in clinical practice</article-title>
          <source>Circulation</source>
          <year>2018</year>
          <volume>138</volume>
          <issue>16</issue>
          <fpage>1623</fpage>
          <lpage>1635</lpage>
        </element-citation>
      </ref>
      <ref id="B36">
        <label>36</label>
        <mixed-citation>36. Nakajima K, Kudo T, Nakata T, Kiso K, Kasai T, Taniguchi Y, et al. Diagnostic accuracy of an artificial neural network compared with statistical quantitation of myocardial perfusion images: a Japanese multicenter study. Eur J Nucl Med Mol Imaging. 2017;44(13):2280-9.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Nakajima</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Kudo</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Nakata</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Kiso</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Kasai</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Taniguchi</surname>
              <given-names>Y</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Diagnostic accuracy of an artificial neural network compared with statistical quantitation of myocardial perfusion images: a Japanese multicenter study</article-title>
          <source>Eur J Nucl Med Mol Imaging</source>
          <year>2017</year>
          <volume>44</volume>
          <issue>13</issue>
          <fpage>2280</fpage>
          <lpage>2289</lpage>
        </element-citation>
      </ref>
      <ref id="B37">
        <label>37</label>
        <mixed-citation>37. Mortazavi BJ, Bucholz EM, Desai NR, Huang C, Curtis JP, Masoudi FA, et al. Comparison of machine learning methods with national cardiovascular data registry models for prediction of risk of bleeding after percutaneous coronary intervention. JAMA Netw Open. 2019;2(7):e196835.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Mortazavi</surname>
              <given-names>BJ</given-names>
            </name>
            <name>
              <surname>Bucholz</surname>
              <given-names>EM</given-names>
            </name>
            <name>
              <surname>Desai</surname>
              <given-names>NR</given-names>
            </name>
            <name>
              <surname>Huang</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Curtis</surname>
              <given-names>JP</given-names>
            </name>
            <name>
              <surname>Masoudi</surname>
              <given-names>FA</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Comparison of machine learning methods with national cardiovascular data registry models for prediction of risk of bleeding after percutaneous coronary intervention</article-title>
          <source>JAMA Netw Open</source>
          <year>2019</year>
          <volume>2</volume>
          <issue>7</issue>
          <elocation-id>e196835</elocation-id>
        </element-citation>
      </ref>
      <ref id="B38">
        <label>38</label>
        <mixed-citation>38. Hernesniemi JA, Mahdiani S, Tynkkynen JA, Lyytik&#228;inen LP, Mishra PP, Lehtim&#228;ki T, et al. Extensive phenotype data and machine learning in prediction of mortality in acute coronary syndrome - the MADDEC study. Ann Med. 2019;51(2):156-63.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Hernesniemi</surname>
              <given-names>JA</given-names>
            </name>
            <name>
              <surname>Mahdiani</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Tynkkynen</surname>
              <given-names>JA</given-names>
            </name>
            <name>
              <surname>Lyytik&#228;inen</surname>
              <given-names>LP</given-names>
            </name>
            <name>
              <surname>Mishra</surname>
              <given-names>PP</given-names>
            </name>
            <name>
              <surname>Lehtim&#228;ki</surname>
              <given-names>T</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Extensive phenotype data and machine learning in prediction of mortality in acute coronary syndrome - the MADDEC study</article-title>
          <source>Ann Med</source>
          <year>2019</year>
          <volume>51</volume>
          <issue>2</issue>
          <fpage>156</fpage>
          <lpage>163</lpage>
        </element-citation>
      </ref>
      <ref id="B39">
        <label>39</label>
        <mixed-citation>39. Nannen V. The Paradox of Overfitting Volker Nannen. [thesis] . Pa&#237;ses Baixos: Faculty of Artificial Intelligence at the University of Groningen; 2003.</mixed-citation>
        <element-citation publication-type="thesis">
          <person-group person-group-type="author">
            <name>
              <surname>Nannen</surname>
              <given-names>V</given-names>
            </name>
          </person-group>
          <source>The Paradox of Overfitting Volker Nannen</source>
          <comment>thesis</comment>
          <publisher-loc>Pa&#237;ses Baixos</publisher-loc>
          <publisher-name>Faculty of Artificial Intelligence at the University of Groningen</publisher-name>
          <year>2003</year>
        </element-citation>
      </ref>
      <ref id="B40">
        <label>40</label>
        <mixed-citation>40. Cortes C, Vapnik V. Support-vector networks. Mach Learn. 1995;20:273-97.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Cortes</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Vapnik</surname>
              <given-names>V</given-names>
            </name>
          </person-group>
          <article-title>Support-vector networks</article-title>
          <source>Mach Learn</source>
          <year>1995</year>
          <volume>20</volume>
          <fpage>273</fpage>
          <lpage>297</lpage>
        </element-citation>
      </ref>
      <ref id="B41">
        <label>41</label>
        <mixed-citation>41. Bayes T. An essay towards solving a problem in the doctrine of chances. By the late Rev. Mr. Bayes, F. R. S. communicated by Mr. Price, in a letter to John Canton, A. M. F. R. S. Philos Trans R Soc Lond. 1763;53:370-418.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Bayes</surname>
              <given-names>T</given-names>
            </name>
          </person-group>
          <article-title>An essay towards solving a problem in the doctrine of chances. By the late Rev. Mr. Bayes, F. R. S. communicated by Mr. Price, in a letter to John Canton, A. M. F. R. S</article-title>
          <source>Philos Trans R Soc Lond</source>
          <year>1763</year>
          <volume>53</volume>
          <fpage>370</fpage>
          <lpage>418</lpage>
        </element-citation>
      </ref>
      <ref id="B42">
        <label>42</label>
        <mixed-citation>42. Webb GI, Boughton JR, Wang Z. Not so naive bayes: aggregating one-dependence estimators. Mach Learn. 2005;58(1):5-24.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Webb</surname>
              <given-names>GI</given-names>
            </name>
            <name>
              <surname>Boughton</surname>
              <given-names>JR</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>Z</given-names>
            </name>
          </person-group>
          <article-title>Not so naive bayes: aggregating one-dependence estimators</article-title>
          <source>Mach Learn</source>
          <year>2005</year>
          <volume>58</volume>
          <issue>1</issue>
          <fpage>5</fpage>
          <lpage>24</lpage>
        </element-citation>
      </ref>
      <ref id="B43">
        <label>43</label>
        <mixed-citation>43. Watkins, DS. Fundamentals of matrix computations. 2th ed. New York: Wiley-Interscience; 2002.</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Watkins</surname>
              <given-names>DS</given-names>
            </name>
          </person-group>
          <source>Fundamentals of matrix computations</source>
          <edition>2th</edition>
          <publisher-loc>New York</publisher-loc>
          <publisher-name>Wiley-Interscience</publisher-name>
          <year>2002</year>
        </element-citation>
      </ref>
      <ref id="B44">
        <label>44</label>
        <mixed-citation>44. Cover T, Hart P. (1967). Nearest neighbor pattern classification. IEEE Trans Inf Theory. 1967;13(1):21-7.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Cover</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Hart</surname>
              <given-names>P.</given-names>
            </name>
          </person-group>
          <year>1967</year>
          <article-title>Nearest neighbor pattern classification</article-title>
          <source>IEEE Trans Inf Theory</source>
          <comment>1967</comment>
          <volume>13</volume>
          <issue>1</issue>
          <fpage>21</fpage>
          <lpage>27</lpage>
        </element-citation>
      </ref>
      <ref id="B45">
        <label>45</label>
        <mixed-citation>45. Fix, E., Hodges, J.L. Discriminatory analysis, nonparametric discrimination: Consistency properties. Technical Report 4, USAF School of Aviation Medicine, Randolph Field, Texas, 1951.</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Fix</surname>
              <given-names>E.</given-names>
            </name>
            <name>
              <surname>Hodges</surname>
              <given-names>J.L.</given-names>
            </name>
          </person-group>
          <source>Discriminatory analysis, nonparametric discrimination: Consistency properties. Technical Report 4, USAF School of Aviation Medicine</source>
          <publisher-name>Randolph Field</publisher-name>
          <publisher-loc>Texas</publisher-loc>
          <year>1951</year>
        </element-citation>
      </ref>
      <ref id="B46">
        <label>46</label>
        <mixed-citation>46. Holland JH. Adaptation in natural and artificial systems. 2th ed. Cambridge, MA: MIT Press; 1992.</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Holland</surname>
              <given-names>JH</given-names>
            </name>
          </person-group>
          <source>Adaptation in natural and artificial systems</source>
          <edition>2th</edition>
          <publisher-loc>Cambridge, MA</publisher-loc>
          <publisher-name>MIT Press</publisher-name>
          <year>1992</year>
        </element-citation>
      </ref>
      <ref id="B47">
        <label>47</label>
        <mixed-citation>47. Breiman L. Random forests. Mach Learn. 2001;45(1):5-32.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Breiman</surname>
              <given-names>L</given-names>
            </name>
          </person-group>
          <article-title>Random forests</article-title>
          <source>Mach Learn</source>
          <year>2001</year>
          <volume>45</volume>
          <issue>1</issue>
          <fpage>5</fpage>
          <lpage>32</lpage>
        </element-citation>
      </ref>
      <ref id="B48">
        <label>48</label>
        <mixed-citation>48. Ho TK. Random decision forests. In: Proceedings of the 3rd International Conference on Document Analysis and Recognition; 1995 Aug 14-16; Montreal. Washington, DC: IEEE Computer Society; 1995. p.278-82.</mixed-citation>
        <element-citation publication-type="confproc">
          <person-group person-group-type="author">
            <name>
              <surname>Ho</surname>
              <given-names>TK</given-names>
            </name>
          </person-group>
          <source>Random decision forests</source>
          <comment>Proceedings of the</comment>
          <conf-name>3rd International Conference on Document Analysis and Recognition</conf-name>
          <conf-date>1995 Aug 14-16</conf-date>
          <conf-loc>Montreal</conf-loc>
          <publisher-loc>Washington, DC</publisher-loc>
          <publisher-name>IEEE Computer Society</publisher-name>
          <year>1995</year>
          <fpage>278</fpage>
          <lpage>282</lpage>
        </element-citation>
      </ref>
      <ref id="B49">
        <label>49</label>
        <mixed-citation>49. MacQueen JB. Some Methods for classification and Analysis of Multivariate Observations. Proc. Fifth Berkeley Symp. on Math. Statist. and Prob. 1967;1:281-97.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>MacQueen</surname>
              <given-names>JB</given-names>
            </name>
          </person-group>
          <article-title>Some Methods for classification and Analysis of Multivariate Observations</article-title>
          <source>Proc. Fifth Berkeley Symp. on Math. Statist. and Prob</source>
          <year>1967</year>
          <volume>1</volume>
          <fpage>281</fpage>
          <lpage>297</lpage>
        </element-citation>
      </ref>
      <ref id="B50">
        <label>50</label>
        <mixed-citation>50. McCulloch WS, Pitts W. A logical calculus of ideas immanent in nervous activity. Bull Math Biophys. 1943;5(4):115-33.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>McCulloch</surname>
              <given-names>WS</given-names>
            </name>
            <name>
              <surname>Pitts</surname>
              <given-names>W</given-names>
            </name>
          </person-group>
          <article-title>A logical calculus of ideas immanent in nervous activity</article-title>
          <source>Bull Math Biophys</source>
          <year>1943</year>
          <volume>5</volume>
          <issue>4</issue>
          <fpage>115</fpage>
          <lpage>133</lpage>
        </element-citation>
      </ref>
      <ref id="B51">
        <label>51</label>
        <mixed-citation>51. Rosenblatt F. The perceptron: a probabilistic model for information storage and organization in the brain. Psychol Rev. 1958;65(6):386-408.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Rosenblatt</surname>
              <given-names>F</given-names>
            </name>
          </person-group>
          <article-title>The perceptron: a probabilistic model for information storage and organization in the brain</article-title>
          <source>Psychol Rev</source>
          <year>1958</year>
          <volume>65</volume>
          <issue>6</issue>
          <fpage>386</fpage>
          <lpage>408</lpage>
        </element-citation>
      </ref>
      <ref id="B52">
        <label>52</label>
        <mixed-citation>52. Broomhead DS, Lowe D. Multivariable functional interpolation and adaptive networks. Complex Syst. 1988;2:321-55.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Broomhead</surname>
              <given-names>DS</given-names>
            </name>
            <name>
              <surname>Lowe</surname>
              <given-names>D</given-names>
            </name>
          </person-group>
          <article-title>Multivariable functional interpolation and adaptive networks</article-title>
          <source>Complex Syst</source>
          <year>1988</year>
          <volume>2</volume>
          <fpage>321</fpage>
          <lpage>355</lpage>
        </element-citation>
      </ref>
      <ref id="B53">
        <label>53</label>
        <mixed-citation>53. Friedman JH. Greedy function approximation: a gradient boosting machine. Ann Stat. 2001;29(5):1189-232.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Friedman</surname>
              <given-names>JH</given-names>
            </name>
          </person-group>
          <article-title>Greedy function approximation: a gradient boosting machine</article-title>
          <source>Ann Stat</source>
          <year>2001</year>
          <volume>29</volume>
          <issue>5</issue>
          <fpage>1189</fpage>
          <lpage>1232</lpage>
        </element-citation>
      </ref>
      <ref id="B54">
        <label>54</label>
        <mixed-citation>54. Friedman JH. Stochastic gradient boosting. Comput. Stat. Data Anal. 2002;38(4):367-78.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Friedman</surname>
              <given-names>JH</given-names>
            </name>
          </person-group>
          <article-title>Stochastic gradient boosting</article-title>
          <source>Comput. Stat. Data Anal</source>
          <year>2002</year>
          <volume>38</volume>
          <issue>4</issue>
          <fpage>367</fpage>
          <lpage>378</lpage>
        </element-citation>
      </ref>
      <ref id="B55">
        <label>55</label>
        <mixed-citation>55. Wang Y, Kosinski M. Deep neural networks are more accurate than humans at detecting sexual orientation from facial images. J Pers Soc Psychol. 2018;114(2):246-57.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Wang</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Kosinski</surname>
              <given-names>M</given-names>
            </name>
          </person-group>
          <article-title>Deep neural networks are more accurate than humans at detecting sexual orientation from facial images</article-title>
          <source>J Pers Soc Psychol</source>
          <year>2018</year>
          <volume>114</volume>
          <issue>2</issue>
          <fpage>246</fpage>
          <lpage>257</lpage>
        </element-citation>
      </ref>
      <ref id="B56">
        <label>56</label>
        <mixed-citation>56. Ma H, Marti-Gutierrez N, Park SW, Wu J, Lee Y, Suzuki K, et al. Correction of a pathogenic gene mutation in human embryos. Nature. 2017;548(7668):413-9.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ma</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Marti-Gutierrez</surname>
              <given-names>N</given-names>
            </name>
            <name>
              <surname>Park</surname>
              <given-names>SW</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Lee</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Suzuki</surname>
              <given-names>K</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Correction of a pathogenic gene mutation in human embryos</article-title>
          <source>Nature</source>
          <year>2017</year>
          <volume>548</volume>
          <issue>7668</issue>
          <fpage>413</fpage>
          <lpage>419</lpage>
        </element-citation>
      </ref>
      <ref id="B57">
        <label>57</label>
        <mixed-citation>57. Shah SJ, Katz DH, Deo RC. Phenotypic spectrum of heart failure with preserved ejection fraction. Heart Fail Clin. 2014;10(3):407-18.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Shah</surname>
              <given-names>SJ</given-names>
            </name>
            <name>
              <surname>Katz</surname>
              <given-names>DH</given-names>
            </name>
            <name>
              <surname>Deo</surname>
              <given-names>RC</given-names>
            </name>
          </person-group>
          <article-title>Phenotypic spectrum of heart failure with preserved ejection fraction</article-title>
          <source>Heart Fail Clin</source>
          <year>2014</year>
          <volume>10</volume>
          <issue>3</issue>
          <fpage>407</fpage>
          <lpage>418</lpage>
        </element-citation>
      </ref>
      <ref id="B58">
        <label>58</label>
        <mixed-citation>58. Medina EL, Mesquita CT, Loques Filho O. Healthcare social networks for patients with cardiovascular diseases and recommendation systems. Int J Cardiovasc Sci. 2016;29(1):80-5.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Medina</surname>
              <given-names>EL</given-names>
            </name>
            <name>
              <surname>Mesquita</surname>
              <given-names>CT</given-names>
            </name>
            <name>
              <surname>Loques</surname>
              <given-names>O</given-names>
              <suffix>Filho</suffix>
            </name>
          </person-group>
          <article-title>Healthcare social networks for patients with cardiovascular diseases and recommendation systems</article-title>
          <source>Int J Cardiovasc Sci</source>
          <year>2016</year>
          <volume>29</volume>
          <issue>1</issue>
          <fpage>80</fpage>
          <lpage>85</lpage>
        </element-citation>
      </ref>
      <ref id="B59">
        <label>59</label>
        <mixed-citation>59. Bittencourt MS. From evidence-based medicine to precision health: using data to personalize care. Arq Bras Cardiol. 2018;111(6):762-3.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Bittencourt</surname>
              <given-names>MS</given-names>
            </name>
          </person-group>
          <article-title>From evidence-based medicine to precision health: using data to personalize care</article-title>
          <source>Arq Bras Cardiol</source>
          <year>2018</year>
          <volume>111</volume>
          <issue>6</issue>
          <fpage>762</fpage>
          <lpage>763</lpage>
        </element-citation>
      </ref>
    </ref-list>
    <fn-group>
      <fn fn-type="other">
        <p>Vincula&#231;&#227;o acad&#234;mica</p>
        <p>Este artigo &#233; parte de tese de Doutorado de Erito Marques de Souza Filho pela Universidade Federal Fluminense.</p>
      </fn>
      <fn fn-type="other">
        <p>Aprova&#231;&#227;o &#233;tica e consentimento informado</p>
        <p>Este artigo n&#227;o cont&#233;m estudos com humanos ou animais realizados por nenhum dos autores.</p>
      </fn>
    </fn-group>
    <fn-group>
      <fn fn-type="supported-by">
        <p>Fontes de financiamento</p>
        <p>O presente estudo n&#227;o teve fontes de financiamento externas.</p>
      </fn>
    </fn-group>
  </back>
  <sub-article article-type="translation" id="TRen" xml:lang="en">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Review Article</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Artificial Intelligence in Cardiology: Concepts, Tools and Challenges - &#8220;The Horse is the One Who Runs, You Must Be the Jockey&#8221;</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0381-3344</contrib-id>
          <name>
            <surname>Souza</surname>
            <given-names>Erito Marques de</given-names>
            <suffix>Filho</suffix>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
          <xref ref-type="aff" rid="aff2001">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Fernandes</surname>
            <given-names>Fernando de Amorim</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Soares</surname>
            <given-names>Celine Lacerda de Abreu</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Seixas</surname>
            <given-names>Flavio Luiz</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Santos</surname>
            <given-names>Alair Augusto Sarmet M.D. dos</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Gismondi</surname>
            <given-names>Ronaldo Altenburg</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Mesquita</surname>
            <given-names>Evandro Tinoco</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-1466-9413</contrib-id>
          <name>
            <surname>Mesquita</surname>
            <given-names>Claudio Tinoco</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
      </contrib-group>
      <aff id="aff1001">
        <label>1</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal Fluminense, Niter&#243;i, RJ &#8211; Brazil</institution>
      </aff>
      <aff id="aff2001">
        <label>2</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal Rural do Rio de Janeiro - Departamento de Tecnologias e Linguagens, Nova Igua&#231;u, RJ &#8211; Brazil</institution>
      </aff>
      <author-notes>
        <corresp id="c01001">Mailing Address: Erito Marques de Souza Filho &#8226; Universidade Federal Fluminense - Departamento de Medicina Cl&#237;nica &#8211; Av. Marques do Paran&#225;, 303. Postal Code 24033-900, Niter&#243;i, RJ &#8211; Brazil E-mail: mederitomarques@gmail.com</corresp>
        <fn fn-type="con">
          <p>Author contributions</p>
          <p>Conception and design of the research: Souza Filho EM, Seixas FL, Santos AASMD, Gismondi RA, Mesquita ET, Mesquita CT; Acquisition of data: Souza Filho EM, Fernandes FA, Soares CLA, Santos AASMD; Analysis and interpretation of the data: Souza Filho EM, Fernandes FA, Soares CLA, Seixas FL, Gismondi RA, Mesquita ET, Mesquita CT; Writing of the manuscript: Souza Filho EM, Fernandes FA, Soares CLA, Gismondi RA, Mesquita ET; Critical revision of the manuscript for intellectual content: Seixas FL, Santos AASMD, Mesquita CT.</p>
        </fn>
        <fn fn-type="conflict">
          <p>Potential Conflict of Interest</p>
          <p>No potential conflict of interest relevant to this article was reported.</p>
        </fn>
      </author-notes>
      <abstract>
        <title>Abstract</title>
        <p>The recent advances at hardware level and the increasing requirement of personalization of care associated with the urgent needs of value creation for the patients has helped Artificial Intelligence (AI) to promote a significant paradigm shift in the most diverse areas of medical knowledge, particularly in Cardiology, for its ability to support decision-making and improve diagnostic and prognostic performance. In this context, the present work does a non-systematic review of the main papers published on AI in Cardiology, focusing on its main applications, potential impacts and challenges.</p>
      </abstract>
      <kwd-group xml:lang="en">
        <kwd>Artificial Intelligence/trends</kwd>
        <kwd>Computer Systems/trends</kwd>
        <kwd>Machine Learning/trends</kwd>
        <kwd>Cardiovascular Diseases</kwd>
        <kwd>Clinical Decision-Making</kwd>
      </kwd-group>
    </front-stub>
    <body>
      <sec sec-type="intro">
        <title>Introduction</title>
        <p>A person&#8217;s everyday life necessitates a huge amount of knowledge about the world and the volume of data in health grows exponentially throughout the world.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> On the other hand, biomedical knowledge is always expanding in an active and dynamic way and cannot be processed or stored by a single human brain. This situation makes it very difficult for the contemporary physician to keep up-to-date with such a broad spectrum of new data and findings, as well as to use such information easily and in a timely manner.<sup><xref ref-type="bibr" rid="B2">2</xref></sup> Adding to this framework are the significant burnout rates among health professionals<sup><xref ref-type="bibr" rid="B3">3</xref> , <xref ref-type="bibr" rid="B4">4</xref></sup> and the important impact of medical errors - which in the United States represent the third leading cause of death.<sup><xref ref-type="bibr" rid="B5">5</xref></sup> This panorama brings with it the need to reorganize the productive structure of health services, associated with various challenges and new perspectives. Given that the current health system is generally unproductive and/or expensive, it is imperative to develop alternative and innovative strategies. The central focus for achieving this goal should be to increase the value for the patient &#8211; outcomes reached per dollar spent &#8211; so that good outcomes, efficiently obtained, are a target to be pursued.<sup><xref ref-type="bibr" rid="B6">6</xref></sup></p>
        <p>Besides, the recent advances at hardware level related to parallel processing, the existence of several machine-learning methods and the huge amount of annotated data contributed for artificial intelligence (AI) to promote a significant paradigm shift in the most diverse areas of medical knowledge and, particularly in Cardiology, for its ability to support decision-making that can improve diagnostic and prognostic performance. These impacts ought to be evaluated from the perspective of patient safety, personalization of care, value creation for the patients, within a scope of technological surveillance &#8211; that gradually consolidates AI as fundamental for a medical practice of excellence.<sup><xref ref-type="bibr" rid="B7">7</xref> - <xref ref-type="bibr" rid="B11">11</xref></sup></p>
        <p>This scenario makes AI, given its importance, be considered by many as the new electricity. The main journals in cardiology have published reviews in this area and the number of articles on the subject follows a growing trend, as shown in <xref ref-type="fig" rid="f01001">Figure 1</xref> &#8211; this behavior is also seen in other medical specialties, such as Neurology. Therefore, the present work performs a non-systematic review of the main papers published on AI in Cardiology, focusing on its main applications, potential impacts and challenges. The next section presents the conceptual fundamentals on the topic, followed by a discussion on why cardiology needs AI and its main tools. Finally, the main challenges, perspectives and conclusions are presented.</p>
        <p>
          <fig id="f01001">
            <label>Figure 1</label>
            <caption>
              <title>&#8211; <italic>Evolution of the number of works relating (Artificial Intelligence or Machine Learning) and Cardiology. Source: Pubmed. Accessed on 12/15/2018. Mesh Words: Cardiology and Machine Learning.</italic></title>
            </caption>
            <graphic xlink:href="1678-4170-abc-114-04-0718-80431-gf01-en.jpg"/></fig>
        </p>
        <sec>
          <title>What is artificial intelligence?</title>
          <p>The term AI was used for the first time at the Dartmouth Conference in 1956.<sup><xref ref-type="bibr" rid="B12">12</xref></sup> Nevertheless, the possibility of machines being able to simulate human behavior and actually think was raised earlier by Alan Turing in 1950, who developed a test in order to differentiate humans from machines &#8211; thus named Turing test.<sup><xref ref-type="bibr" rid="B13">13</xref></sup></p>
          <p>Basically, AI is the product of the combination of sophisticated mathematical models and computation, which allows the development of complex algorithms capable of emulating human intelligence. All this process starts with the construction of a database representative of the problem that one wishes to study &#8211; adequately collected and processed&#8211; called healthy data. This step is of fundamental importance, as the algorithms will probably not perform well if this prerequisite is not obtained: " <italic>garbage in, garbage out</italic> ".</p>
          <p>The nature of these data is quite varied, ranging from socio-environmental, clinical-laboratory, omic-data (e.g., metabolome, proteome, epigenome, lipidome) to information on red, green and blue intensities (RGB system) of each pixel that composes an image, for example. Equally diversified sources of such data include those obtained from electronic medical records or even wearable devices. In this context, the term Big Data is used to describe a huge collection of data for which traditional methods of analysis are unsuccessful in analyzing, searching, interpreting and storing.<sup><xref ref-type="bibr" rid="B9">9</xref></sup></p>
          <p>We highlight the use of these tools in problems of classification, regression, and clusterization. After obtaining healthy data and building the database, it is important to evaluate which mathematical models of AI are most appropriate for the problem that one wishes to solve. Then, the chosen models must be implemented using some programming language. A combination of models can also be useful. The results obtained by the algorithm should be analyzed in terms of both the coherence and suitability. These steps are summarized in <xref ref-type="fig" rid="f02001">Figure 2</xref> .</p>
          <p>
            <fig id="f02001">
              <label>Figure 2</label>
              <caption>
                <title>&#8211; <italic>Main Illustration.</italic></title>
              </caption>
              <graphic xlink:href="1678-4170-abc-114-04-0718-80431-gf02-en.jpg"/></fig>
          </p>
        </sec>
        <sec>
          <title>Why does cardiology need artificial intelligence?</title>
          <p>The development of AI algorithms has the advantage of not requiring many assumptions in relation to underlying data.<sup><xref ref-type="bibr" rid="B8">8</xref></sup> Another point is that the nature of these mathematical-computational models allows, from observational data, a high level of evidence due to its high performance, which certainly represents a significant paradigm shift in evidence-based medicine. It should be noted that traditional clinical trials are generally slow, expensive, time-consuming, and limited in size.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> In addition, when the database is fed with more (healthy) data, in general, there is an improvement in the performance of the algorithms &#8211; which allows the studies to have a continuous character over time.</p>
          <p>This new archetype can guide the allocation of scarce resources in the health area and facilitate the efficient and accurate identification of decisions that favor the individualization of care based on the flow of information that emerge from an integrated and complex ecosystem: it is a precision medicine.<sup><xref ref-type="bibr" rid="B15">15</xref> , <xref ref-type="bibr" rid="B16">16</xref></sup> Therefore, it can be inferred that the practice of the cardiovascular sciences will have significant impacts, which will translate into a personalized approach and improved outcomes.</p>
        </sec>
        <sec>
          <title>Basic concepts in artificial intelligence</title>
          <p>A generic database can be arranged in a matrix of rows and columns. Each line denotes an element from a set of objects to be evaluated according to the same features. Each column, in turn, expresses the values of a given attribute for the various rows in the database and each line represents a lesson to be learned by the mathematical-computational model. In this way, the term Machine Learning (ML) brings with it a possibility of " <italic>learning</italic> " from a set of lessons. The term AI is often used interchangeably with the term ML. However, ML is a subset of AI algorithms related the ability of learn from a large amount of data. AI is wider and encompass performing tasks that are normally related to human intelligence such as pattern recognition, problem solving, understanding language or recognizing objects and sounds.<sup><xref ref-type="bibr" rid="B17">17</xref></sup></p>
          <p>It is often said that the types of learning can be:</p>
          <list list-type="alpha-lower">
            <list-item>
              <p><bold>Supervised:</bold> when the algorithm receives information about each lesson as well as the labels associated with it, having an important role in relation to the prediction. For example, if it is desired to predict whether a patient is more susceptible to cough with the use of angiotensin-converting enzyme inhibitors, analysis should be performed based on a healthy database containing a group of patients that showed such a reaction and another group in which this fact was not observed.</p>
            </list-item>
            <list-item>
              <p><bold>Unsupervised:</bold> when the lesson labels are not provided <italic>a priori</italic> , it is up to the algorithm to find hidden structures in the database. A hypothetical example is the clusterization of a database of patients with hypertrophic cardiomyopathy according to imaging findings.</p>
            </list-item>
            <list-item>
              <p><bold>Reinforcement:</bold> inspired by behavioral biology, it is a kind of reward-based learning.<sup><xref ref-type="bibr" rid="B18">18</xref> , <xref ref-type="bibr" rid="B19">19</xref></sup></p>
            </list-item>
          </list>
          <p>Another important concept is that of cognitive computing. It can be understood as a set of self-learning systems intended to imitate the human thought process based on the use of ML tools, pattern recognition and natural language processing.<sup><xref ref-type="bibr" rid="B9">9</xref></sup> IBM Watson is an example of cognitive computing in the medical field.<sup><xref ref-type="bibr" rid="B20">20</xref> , <xref ref-type="bibr" rid="B21">21</xref></sup></p>
        </sec>
        <sec>
          <title>Some artificial intelligence tools and applications</title>
          <p>Currently, there is a multiplicity of models of ML each of them with diverse particularities, varied uses and limitations. The applications of some of these models in Cardiology are explained in the following paragraphs, while a brief description of each of them and their type is shown in <xref ref-type="table" rid="t1001">Table 1</xref> .</p>
          <p>
            <table-wrap id="t1001">
              <label>Table 1</label>
              <caption>
                <title>&#8211; Brief description and classification of the main ML tools</title>
              </caption>
              <table frame="hsides" rules="groups">
                <colgroup width="33%">
                  <col/>
                  <col/>
                  <col/>
                </colgroup>
                <thead>
                  <tr>
                    <th align="left">Tool</th>
                    <th>Description</th>
                    <th>Learning</th>
                  </tr>
                </thead>
                <tbody>
                  <tr>
                    <td>SVM</td>
                    <td align="center">It is useful for two-group classification problems. The idea is to find a function called hyperplane from the resolution of a linear system built from the various lessons of the training subset. <xref ref-type="bibr" rid="B40"><sup>40</sup></xref> This hyperplane is used to cluster the lessons of the test subset into two disjoint groups.</td>
                    <td align="center">Supervised</td>
                  </tr>
                  <tr>
                    <td>NB</td>
                    <td align="center">It was inspired in the studies of the reverend Bayes on conditional probability. <xref ref-type="bibr" rid="B41"><sup>41</sup></xref> These probabilities are used to identify the category (out of a total n possible) that a particular lesson belongs to. <xref ref-type="bibr" rid="B42"><sup>42</sup></xref>
										</td>
                    <td align="center">Supervised</td>
                  </tr>
                  <tr>
                    <td>KNN</td>
                    <td align="center">It is said that a vector norm is a mathematical function, which satisfies specific properties, and associates a vector with a value greater than or equal to zero. <xref ref-type="bibr" rid="B43"><sup>43</sup></xref> The norm of the difference between two vectors is the distance between them. The KNN uses a norm to calculate the distance between all the vectors (lessons) that make up the database. Then, for each vector of the database, the k vectors closest to it are determined. The inclusion in a given group is obtained from a majority voting system among the neighbors. <xref ref-type="bibr" rid="B44"><sup>44</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B45"><sup>45</sup></xref>
										</td>
                    <td align="center">Supervised</td>
                  </tr>
                  <tr>
                    <td>AG</td>
                    <td align="center">Algorithms inspired by the biological evolution of species, in which each possible candidate to solve the problem is modeled as a chromosome consisting of a set of genes, which during the execution of the algorithm undergoes operations of crossing-over and mutation in order to obtain better solutions than the current ones. <xref ref-type="bibr" rid="B46"><sup>46</sup></xref> This way, they allow a database to be separated, for example, into two distinct groups &#8211; which have or do not have a particular characteristic.</td>
                    <td align="center">Supervised</td>
                  </tr>
                  <tr>
                    <td>RF</td>
                    <td align="center">This method is based on the construction of several decision trees. The first step is to get several random samples (with reposition) of lessons to build other databases, a process that is called bootstrapping. Each of these new databases will give rise to a decision tree, which is obtained iteratively, from a subset of variables (features). After the construction of all trees, a new lesson in the database should be allocated to the group that has the largest number of decision trees, showing that it belongs to this group (majority of votes). <xref ref-type="bibr" rid="B47"><sup>47</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B48"><sup>48</sup></xref>
										</td>
                    <td align="center">Supervised</td>
                  </tr>
                  <tr>
                    <td>K-means</td>
                    <td align="center">It allows partitioning a database into k groups with similar characteristics. To do so, it is necessary to update, in an iterative way, a set of vectors, called reference centroids of each group and to calculate the distance of each lesson to each one. A lesson is always allocated to the centroid for which it has the shortest distance. The elbow chart is generally used to determine the ideal number of groups to separate from the database. <xref ref-type="bibr" rid="B49"><sup>49</sup></xref>
										</td>
                    <td align="center">Unsupervised</td>
                  </tr>
                  <tr>
                    <td>ANN</td>
                    <td align="center">Inspired in biological nerve systems, a structure called a graph - a set of nodes and edges - is used in which nodes are layered and connected by valued edges, which represent a weight assigned to a given connection. The idea is that from a set of inputs, these weights are used properly to produce an output. Several architectures have been proposed for neural networks, from simpler ones such as the perceptron, to more sophisticated ones, such as the radial basis function, convolutional networks and deep learning. In deep learning, in addition to the input and output layers, there are hidden layers that increase significantly the number of weights to be updated and often require huge computational efforts. Convolutional network is a type of deep leaning inspired in visual cortex of animals that have an important role in image analysis. Autoencoders and Kohonen neural networks are examples of unsupervised learning. <xref ref-type="bibr" rid="B1"><sup>1</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B7"><sup>7</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B50"><sup>50</sup></xref><sup>-</sup><xref ref-type="bibr" rid="B52"><sup>52</sup></xref>
										</td>
                    <td align="center">Unsupervised or Supervised</td>
                  </tr>
                  <tr>
                    <td>GB</td>
                    <td align="center">It is a tree-based method that uses gradient, vectors related to the direction of maximum increase in a math function, to produce sequential decision trees to be combined to brush up on the prediction. Variants of this approach include Stochastic Gradient Descent that incorporates a random subsampling to GB. <xref ref-type="bibr" rid="B53"><sup>53</sup></xref><sup>,</sup><xref ref-type="bibr" rid="B54"><sup>54</sup></xref>
										</td>
                    <td align="center">Supervised</td>
                  </tr>
                </tbody>
              </table>
            </table-wrap>
          </p>
          <list list-type="alpha-lower">
            <list-item>
              <p><bold>Support Vector Machine (SVM):</bold> used by Samad et al.,<sup><xref ref-type="bibr" rid="B22">22</xref></sup> to predict with success the deterioration of ventricular function in patients with repaired tetralogy of Fallot from a database of 153 patients with clinical, electrocardiographic and cardiac magnetic resonance imaging data. In relation to predicting any deterioration (minor or major) vs. no deterioration, the mean area under the curve (AUC) was 0.82 &#177; 0.06.<sup><xref ref-type="bibr" rid="B22">22</xref></sup> Berikol et al.<sup><xref ref-type="bibr" rid="B23">23</xref></sup> used clinical, laboratory (troponin I and CK-MB levels), ECG, and echocardiographic data from 228 patients who presented at the emergency department with chest pain for classification regarding the presence or absence of Acute Coronary Syndrome. Accuracy, sensitivity and specificity were, respectively, 99.19, 98.22 and 100%.<sup><xref ref-type="bibr" rid="B23">23</xref></sup> Betancur et al.<sup><xref ref-type="bibr" rid="B24">24</xref></sup> also used SVM to more precisely define mitral valve plane (VP) positioning during left ventricular segmentation in Single-Photon Emission Computed Tomography (SPECT) exams. Images of 392 patients were analyzed and the good results obtained were compatible with the opinion of experts in the area &#8211; AUC: 0.82 [0.74-0.9] for regional detection of obstructive stenosis and ischemic total perfusion deficit areas.<sup><xref ref-type="bibr" rid="B24">24</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>Naive Bayes (NB):</bold> Paredes et al.,<sup><xref ref-type="bibr" rid="B25">25</xref></sup> used an NB fusion and genetic algorithm to predict the risk of occurrence of cardiovascular events (e.g., hospitalization or death) based on data from 559 Acute Coronary Syndrome-Non-ST Segment Myocardial Infarction (ACS-NSTEMI) patients. Sensitivity and specificity were, respectively, 79.8, 83.8.<sup><xref ref-type="bibr" rid="B25">25</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>K-nearest neighbors (KNN):</bold> Al-Mallah et al.<sup><xref ref-type="bibr" rid="B26">26</xref></sup> compared the prediction of all-cause mortality in 10 years between the classical logistic regression model and the KNN, considering a database of 34,212 patients with clinical information and information obtained after the treadmill test using the standard protocol of Bruce.<sup><xref ref-type="bibr" rid="B26">26</xref></sup> The results obtained by this ML tool showed a sensitivity of 87.4% and specificity of 97.2%, better than the predictive performance of the traditional Atherosclerosis Cardiovascular Disease Risk Score (ASCVD).</p>
            </list-item>
            <list-item>
              <p><bold>Genetic algorithms (GA):</bold> Smisek et al.<sup><xref ref-type="bibr" rid="B27">27</xref></sup> developed a wearable device to detect arrhythmias from the information record of a single-lead electrocardiogram. The data were analyzed from a combination of the (SVM), decision tree and threshold-based rules. Genetic algorithms were used to select the most appropriate characteristics to be used in the work. In relation to the detection of atrial fibrillation, an F1 score (harmonic mean of positive predictive value and sensitivity) of 0.81 was obtained.<sup><xref ref-type="bibr" rid="B27">27</xref></sup> Stuckey et al.,<sup><xref ref-type="bibr" rid="B28">28</xref></sup> used the Cardiac Phase Space Tomography Analysis - a pioneering method that dispenses with the use of radiation and contrast, as well as performing exercises or pharmacological stress - combined with ML models (e.g., genetic algorithms) to analyze the thoracic phase signals. In this study, the authors used this tool to evaluate patients with coronary disease and chest pain who were referred by the physician for angiography. 606 patients were studied, and the results showed sensitivity of 92%, specificity of 62% and predictive value of 96% for coronary disease.<sup><xref ref-type="bibr" rid="B28">28</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>Random Forests (RF):</bold> Samad et al.,<sup><xref ref-type="bibr" rid="B29">29</xref></sup> analyzed a database consisting of clinical and electrocardiographic variables to evaluate survival in 10 different periods of time (ranging from 6 to 60 months), considering a total of 171,510 patients. RF was used, with excellent results, better than those obtained through traditional scores such as the Framingham risk score and ACC/AHA guideline score. The area under the curve (AUC) was superior to 0.82.<sup><xref ref-type="bibr" rid="B29">29</xref></sup> Ambale-Venkatesh et al.<sup><xref ref-type="bibr" rid="B30">30</xref></sup> used information from noninvasive tests, questionnaires, biomarkers and imaging tests from 6,814 patients to construct 739 variables (features) in order to apply a variant of RF &#8211; called survivor random forests<sup><xref ref-type="bibr" rid="B31">31</xref></sup> &#8211; for predicting cardiovascular events (all-cause death, stroke, all cardiovascular disease, coronary heart disease, atrial fibrillation and heart failure), having performed better than established risk scores, e.g., MESA-CHD, AHA/ASCVD and Framingham, with increased prediction accuracy (decreased Brier score by 10%-25%).<sup><xref ref-type="bibr" rid="B30">30</xref> , <xref ref-type="bibr" rid="B31">31</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>K-means:</bold> Cikes et al.<sup><xref ref-type="bibr" rid="B32">32</xref></sup> used a database consisting of clinical variables and echocardiographic parameters for which two models of ML, Kmeans and Multiple Kernel Learning were applied, in order to categorize the patients into mutually exclusive groups to evaluate the response to resynchronization therapy cardiac. A total of 1,106 patients were analyzed and four disjoint groups were identified, two of them with the best response to therapy.<sup><xref ref-type="bibr" rid="B32">32</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>Artificial Neural Networks (ANN):</bold> Kwon et al.,<sup><xref ref-type="bibr" rid="B33">33</xref></sup> in a multicenter study of 52,131 patients, constructed a deep learning-based early warning system capable of predicting the occurrence of cardiac arrest in a hospital. The model showed high performance when compared to traditional track-and-trigger systems. The area under the curve was 0.82.<sup><xref ref-type="bibr" rid="B33">33</xref></sup> Rubin et al.,<sup><xref ref-type="bibr" rid="B34">34</xref></sup> had promising preliminary results with the use of neural networks with convolutional architecture to evaluate electrocardiographic signs and to classify them in atrial fibrillation, sinus rhythm (normal) or noise &#8211; the F1 score achieved was 0.82.<sup><xref ref-type="bibr" rid="B34">34</xref></sup> Zhang et al.<sup><xref ref-type="bibr" rid="B35">35</xref></sup> also used convolutional neural networks to analyze a database with 14,035 echocardiographic exams to detect the presence of diseases such as hypertrophic cardiomyopathy, cardiac amyloidosis and pulmonary arterial hypertension with a high performance: C statistics were respectively, 0.93, 0.87, and 0.85.<sup><xref ref-type="bibr" rid="B35">35</xref></sup> Nakajima et al.<sup><xref ref-type="bibr" rid="B36">36</xref></sup> used an ANN to evaluate the presence of coronary disease after performing myocardial scintigraphy. Results were obtained with high accuracy and superior performance to the traditional scores used. For example, the AUC for patients with old myocardial infarction based on defects in rest stage was 0.97.<sup><xref ref-type="bibr" rid="B36">36</xref></sup></p>
            </list-item>
            <list-item>
              <p><bold>Gradient Boosting (GB):</bold> Mortazavi et al.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> used GB for prediction of risk of bleeding after percutaneous coronary intervention and demonstrated that these tools can help to identify patients who would benefit from strategies aiming to reduce the bleeding risk. A total of 3,316,465 procedures were analyzed and a C statistic of 0.82 was obtained.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> Hernesniemi et al.,<sup><xref ref-type="bibr" rid="B38">38</xref></sup> also proposed a GB to predict mortality in acute coronary syndrome, analyzing 9,066 consecutive patients. The AUC was 0.89 and the model performed better than GRACE traditional score.<sup><xref ref-type="bibr" rid="B38">38</xref></sup></p>
            </list-item>
          </list>
          <p>It is important to note that when using any ML model, one should keep in mind a major problem that may arise, called overfitting. It occurs when a model describes the examples very well (training subset) and performs poorly when applied to other instances of the same phenomenon.<sup><xref ref-type="bibr" rid="B39">39</xref></sup> In addition, it is worth saying that there is no theoretical result that ensures that any of the AI algorithms is better than the others in any application. Thus, this choice depends on several variables, such as the nature of the problem under analysis, the time and resources available to solve the problem. The combination of techniques generating hybrid models can also be of great value. On the other hand, the use of tools for parallel processing, such as the Graphic Processing Unit (GPU), has been of great value in improving the performance of ML models, especially in relation to computational time needed to run them.</p>
        </sec>
        <sec>
          <title>Challenges and future prospects</title>
          <p>As previously highlighted, AI applications in cardiology have increased greatly in recent years and their growth potential is enormous. However, this scenario brings with it the need to overcome some challenges, such as: ethical limits of use (misuse), improvement of mathematical knowledge, acquisition of healthy data, development of security, need for collaboration, attention to errors and data-based care. All of this is discussed below and is summarized in <xref ref-type="fig" rid="f02001">Figure 2</xref> .</p>
          <list list-type="alpha-lower">
            <list-item>
              <p><bold>Challenge 1 &#8211; ethical limits of use (misuse):</bold> like all disruptive technology, the limits of ethics need to be rethought and widely discussed. ML algorithms can be misused and misleading. As an example, a work of great repercussion was published by Wang and Kosinski (2018). The authors used deep learning and obtained expressive results in the prediction of whether an individual is gay or not from a database of images of the study participants&#8217; faces.<sup><xref ref-type="bibr" rid="B55">55</xref></sup> Similarly, the same AI algorithms can be used to detect, for example, whether or not a patient will develop atrial fibrillation or any future cardiomyopathy. Could this information be used by companies to increase the amounts of their health plans or even deny membership to the plan due to a high cost? What if it is detected that a baby will be born with congenital heart disease due to the analysis of the genetic, clinical-laboratory and image (or other) data of its parents? This could open space for a kind of neoeugenia. This debate has gained an additional emphasis with the emergence of the CRISP-Cas9 technique, which allows DNA editing.<sup><xref ref-type="bibr" rid="B56">56</xref></sup> In this context, by stimulating a debate with society on the subject, transparency and regulation are fundamental pillars to be preserved.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 2</bold><bold>&#8211;</bold><bold>improve math knowledge:</bold> the advent of this new kind of unbelievable human being ( <italic>Homo incredibile</italic> ), which supports its decisions in data, carries with it the fundamental role of mathematics and computation in this currently ongoing revolution. This revolution will bring unimaginable possibilities in medical practice, such as the construction of quality phenomappings - ML models developed with the aim of clustering patients in function of their large mass of phenotypic characteristics in order to facilitate the decision-making process.<sup><xref ref-type="bibr" rid="B57">57</xref></sup> Thus, it is necessary that these competencies are stimulated early, mainly with a focus on solving problems related to the reality for which one wishes to promote improvements. This will certainly be reflected in a need to reformulate the cardiovascular contents (and why not say, medical content in general) of undergraduate and postgraduate courses in Medicine: a passive or merely expositive education, with an extensive load and that prioritizes the capacity of the student's memory seems, more and more, to be inadequate, as one realizes that Medicine must be a space for creativity and value generation.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 3 &#8211;</bold> get healthy data: the use of healthy data is of fundamental value for the success of the algorithms. Thus, it is required that health units encourage their health professionals regarding the thoroughness at the level of data filling/obtaining as well as maintaining any data sources, from forms, electronic medical records, image data or even unconventional data, such as those obtained by Medina et al.<sup><xref ref-type="bibr" rid="B58">58</xref></sup> - who developed a successful Online Social Networks Health tool in which the patient himself anonymously inserts health monitoring information, including physiological data, daily activities, emotional states, and interaction with others patients.<sup><xref ref-type="bibr" rid="B58">58</xref></sup> Therefore, data management becomes as important as other routine behaviors in evidence-based medicine, such as proper handwashing or even the use of a defibrillator during cardiac arrest. In this way, the formation of multidisciplinary data teams and the constant training of the teams assume a primordial role. It is noteworthy that much of the slowness and difficulty that some health units have in using ML models is tied to absent or incipient healthy data.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 4 &#8211;</bold> get security: the advent of these tools brings with it a fundamental concern with data security, to a level never before experienced, as access to such data by unauthorized persons can lead to catastrophic consequences for both health institutions and the patients. The creation of a security team plays an important role in this new process. The General Data Protection Regulation represents an advance in this direction. Blockchain and its variants are important tools that can improve security substantially.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 5 &#8211;</bold> need for collaboration (the African proverb): there is an African proverb that says, " <italic>if you want to go fast go by yourself, but if you want to go far go with many</italic> ". This applies a lot to this data environment: collaboration between institutions allows the construction of huge healthy databases (Big Data), which tends to favor the performance of ML algorithms.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 6 &#8211;</bold> dealing with errors: one important issue concerns the errors of AI models. It is inadequate to believe that such models are error-free. It may, for example, be the result of overfitting or occurring by using unhealthy data - which make the results unreliable. However, the practice has shown high performance in several applications. These models are probabilistic, and it is always desirable that their errors be minimal. This scenario has clinical implications, for example, an AI model that predicts with 99% probability that a patient has a greater propensity than the general population to have cardiac myocarditis or amyloidosis. There is a probability, although small, that this will not occur, and that the procedure adopted by the cardiologist is inadequate. In that case, the question is who can be held accountable in these cases? Is it appropriate? Should the patient sign a consent form in these cases? Certainly, the solution includes robust regulation of the use of these tools and strengthening of a new type of relationship: physician-patient-data.</p>
            </list-item>
            <list-item>
              <p><bold>Challenge 7 &#8211;</bold> data-based care management: while ML's tools follow an inexorable path, on the other hand, several healthcare professionals remain fearful about these tools because of its possible ability to replace physicians in their tasks. However, when the history of Medicine is remembered, it is worth mentioning, for example, that the appearance of automated machines to perform the whole blood count did not replace the hematologist, but rather resulted in a greater speed of the work process and allowed the professional to be able to act in other important issues in the specialty.</p>
            </list-item>
          </list>
          <p>The central idea is to provide better support for decision-making, including better performance. It is data-driven care management with a high dynamism and constant updating - which will promote greater personalization of care<sup><xref ref-type="bibr" rid="B59">59</xref></sup> and a real-time evaluation of the experience of the health system users, aiming at generating value for the patient. In this context, the mechanical tasks will be substitutable and a diversity of new tasks will be included into the routine of the cardiologist of precision, from the adequate construction of the databases to the critical reflection on the results obtained by the mathematical-computational models, as well as the development of an adequate physician-patient-data relationship. Therefore, there is a migration of human skills as well as the expansion of their capabilities from the emergence of new tools, which should be part of the technical arsenal of the 21<sup>st</sup>century cardiologist. This panorama allows us to compare ML models to a horse and doctors to jockeys: " <italic>the horse is the one who runs , you must be the jockey</italic> ".</p>
        </sec>
      </sec>
      <sec sec-type="conclusions">
        <title>Conclusions</title>
        <p>AI, in fact, has been shown to be a fundamental tool for the clinical practice of current cardiology. Several applications have been successfully performed and have allowed significant improvements from a diagnostic and therapeutic point of view and in relation to personalized care. To be able to use such tools, it is imperative that healthy data be used, which certainly implies a new design in the modus operandi of many health services. The nature of these data is varied and includes new sources, such as wearable devices and omic-data. On the other hand, this new digital ecosystem requires an acquisition of knowledge not traditionally found in regular medical courses. Therefore, a curricular redesign is required and ought to be object of a profound debate and specific actions.</p>
        <p>On the other hand, the entire panacea brought by AI is not free from challenges such as: the ethics limits of its use, the necessity of improving math knowledge, the building of an ecosystem that ensure high levels of security and confidentiality for the patients, the acquisition of healthy data, the needs of expand the physician-patient-data association, the necessity of collaboration and the data-based care management. In this context, the cardiologist-jockey (or physicians in general) must be a protagonist of changes and has to replace an eventual fear of the tools by a greater involvement with the objective of generating value for the care. It is important to keep in mind possible challenges and obstacles to be overcome and to maintain an engagement and critical sense in the search for solutions: "the horse is the one who runs, you must be the jockey".</p>
      </sec>
    </body>
    <back>
      <fn-group>
        <fn fn-type="other">
          <p>Study Association</p>
          <p>This article is part of the thesis of Doctoral submitted by Erito Marques de Souza Filho, from Universidade Federal Fluminense.</p>
        </fn>
        <fn fn-type="other">
          <p>Ethics approval and consent to participate</p>
          <p>This article does not contain any studies with human participants or animals performed by any of the authors.</p>
        </fn>
      </fn-group>
      <fn-group>
        <fn fn-type="supported-by">
          <p>Sources of Funding</p>
          <p>There were no external funding sources for this study.</p>
        </fn>
      </fn-group>
    </back>
  </sub-article>
</article>
